university of nigeria enugu campus … tochukwu.t_08...faculty of business administration department...

UNIVERSITY OF NIGERIA ENUGU CAMPUS

SCHOOL OF POST GRADUATE STUDIES

FACULTY OF BUSINESS ADMINISTRATION

DEPARTMENT OF MANAGEMENT

TOPIC:

TABLEOFCONTENTS

CHAPTER ONE

Introduction - - - - - - - - 1

CHAPTER TWO

2.1 Observation - - - - - - - 3

2.2 Interviewing - - .. - - - - 5

2.4 Questionnaire - - - - - - - 7

2.6 Measuring - - - - - - - 9

CHAPTER THREE

3.1 Probability sampling techniques - - - 11

3.1.1 Systematic Random Sampling Techniques

3.1.2 - - Cluster Random Sampling Techniques

3.2 Non-probability sampling techniques -

3 -2.1 Convenience Sampling - - - - b

3.2.2Volunteer Sampling - - - -

3.2.3 Judgment Sampling Technique - -

3.2.4 Quote Sampling Technique - - -

CHAPTER FOUR

CHAPTER ONE

INTRODUCTION

There are many ways in which data can be gathered. The

researcher design instrument for a specific research, which

could form a kind of primary source of information if it is

uniquely designed, and it may be used for other study. The

instrument that is developed is based on original information.

Thus, it must have to be subjected to rigorous test such as

validity or reliability. More so, the variables are directly

observed, controlled or manipulated so as to generate date for

the study. When the study is completed, this design still

serves as a guide for other research studies.

For the purpose of this topic, we shall take a broad view

of the terms "data", "techniques" and social management

science as a way of helping to understand and appreciate the

importance of their interrelationship in today's society and

management.

Data are group of raw random symbols such as records

of day-to-day figure values activities of an organization.

Examples of day-to-day records that can be kept are records

creditors of the organization, pay roll, purchases made, which

presents events and things that have taken place.

For processing purposes, data are organized into

structures, fields and data base the analysis collects data,

organizes, analyses and processes them into relevant

information that can be assessed for the management used in

their decision making. When these data are gathered and

processed they become information, which is useful for

management decision-making,

Data can be obtained through reading, observation,

country, measuring, weighing, questionnaire, sampling Test,

Interview etc.

There are two main source of data gathering:

The primary and secondary sources.

An actual eyewitness or a participant supplies the

primary sources. They are obtained from the representative

samples by means of direct observation of events, etc.

The secondary source of data include all types of

information from the primary sources but this time, they are

recorded, reported or supplied by someone else. b

Techniques are methods the researcher (analyst) uses to

accomplish his task after data have been gathered.

CHAPTER TWO

This chapter deals on methods of data gathering and the

advantages/disadvantages of each method. The methods are

as follows:

Interview

Reading

Questionnaire

Sampling and

Measuring

Test

2.1 Observation

This should be the best way of obtaining data about what

actually happen because people could be saying one thing

duringinterview but in practices a different thing is done.

Unfortunately, the use of observation for data gathering b

has a major draw basic, which is workers in the system may

consciously select what they want to do as long as the

observation lasts. It entails watching people carry out their

various tasks in the normal way. This method of data

gathering is time consuming on the park of the observer.

Advantages of Observation

It records the actual events rather than people's

opinions.

An articulated researcher will not usually inform his

observed group so as to avoid biases through the

creation of an artificial situation.

It is easy and less expensive unlike the interview

Disadvantages of Observation

Observation is based on current events or happenings.

It does not reveal the past except from secondary

sources.

Highly skilled and psychologically inclined observer is

needed. b

3) Most time, observations made not be valid due to

human indifference, attitude to issues at times or even

due to pretence from the subjects involved in the

study.

2.2 Interviewing

Interviewing is probably the most valuable method, if the

interviewer interviews the right people in the right order and

with right questions. Success in an interview is as dependent

upon the preparation made for US on the way in which is

conducted. Therefore adequate preparation should be made

before embarking on Interview.

Advantages

It gives room for more information to the collected this

is based on the interviewer's self-disposition.

The Interviewer, who has a psychological attitude or

background, is able to deduce when information is

faulty or not form.

Skillful interviewers made uninterested respondent

willing to answer the questions, since it can be

conducted face-to-face-, the interview can easily

explain in his questions tactfully on the respondent.

4) A high response rate particularly when the issue is

sensitive, that is relating to life or self-development

issue.

Disadvantages of interviews

1) The method is quite expensive.

2) Some respondents may not reveal confidential

information.

3) Some degree of biases could be unknowingly

introduced when the interviewer is making judgment.

4) Information collected is not qualifiedly, thus making

analysis a bit difficult.

2.3 Reading

This method covers reading of appropriate journals,

procedure manuals, company reports both internal reports

and published annual reports, staff hand books and

department records.

Advantages

1) It is easy and less expensive.

2) It gives room for more information to be collected

3) It does not required highly skilled and

4) Psychologically include researcher.

Disadvantages of Reading

1) It is difficulty to deduce when data /information is

fully or not true.

2 ) It is based on past events and does not reveal the

current events or happening.

2.4 Questionnaire

This method of data gathering should be least use for

data gathering because of its inherent constraints. The

constraints include the difficulty of producing questions that

are intelligible and unambiguous to reader. Most often the

readers fails to respond to questions not because they do not

want to do so, but because the questions are vague and

confusing, some times contradictory. At best, the

questionnaires are used to find a relatively small amount of

information from a relatively large number of people or events

especially where the distance poses some problems.

Advantages of Questionnaire

The respondent is influenced by the researcher choice

of wards to use in answering the question.

The questions can be answered in considerable dept.

The respondent is spared the potential frustration of

being confined to pre-stated set of response

alternatives.

2.5 Sampling

This is a process whereby a sample is drawn from the

entire kroup usually known as the population, because the

population is too large to be measured completely, among b

other reasons. This sample so drawn is measured in the hope

that the out come is representative of the population. In

dealing with sampling theory what should monitor is the

standard deviation of the sample since other factors or

parameters can be pre-decided. If the standard deviation

varies from one sample to another, it is then necessary to take

samples of increasing size until the standard deviation settle

down to a steady value.

2.6 Measuring

This is not often used except when the data cannot be got

through other means. Usually data that related to times, ratio,

quantities are measured using any other available tools for

measurement, both mechanical and electronic equipment. The

results are normally recorded as mean, mode, median or the

spread that is, standard deviation.

2.7 Test

This is commonly used in experimental research where

the researcher subjects the respondent to a test condition by

ansGering some technically drafted and asking questions,

which have been validly constructed. It tests the respondent's

performance on a situation or determines the level of

achievement.

Advantages of Test

1) It has the last respondent bias

2 ) It is the most fruitful and reliable method.

Disadvantages

It is costly to conduct

It is difficult (very rigorous) t design and execute.

CHAPTER THREE

Data technique is any device constructed for recording or

generating of data. It is the means for generating and

collecting the pertinent Information to be used for solving the

research problem. Adequate and reliable data collected

contribute a lot to knowledge as well as solution of real life

problems. A researcher should arm himself with adequate

courage, pertinence, native wisdom, in order to achieve a

better result.

The data techniques are as follows

(1) Sampling Techniques

a. Simple Random Sampling

b. Stratified random Sampling

c. Systematic Random Sampling

d. Custer Random Sampling

A probability sampling techniques is one in which the

selection of sampling is based on chance, and for which the

change of each units being selected is known. The process of

selecting of probability samples must be stable and

independent.

The variants of probability samples include:

In simple Random Sampling

In simple random sampling, each member (element) of

the population has an equal chance of being selected. Suppose

10 students are to be selected from a class of 100 students.

Each of them would have 10 chances out of 100 or (10%) of

being selected.

Stratified Random Sampling

In this type of sampling technique, the population from

which samples are to be taken is divided into groups called

(strata) on the bases of common characteristics (parameter)

among each group.

3.1.1 Systematic Random Sampling Techniques

This is also known as equal interval sampling, this

technique involves taking every Kth Unit or element in the

form, after picking a random starting point. E.g. 150 route-

card carrying retailers are to be systematically selected from a

list of 12000 for research per posed. The K (periodically

interval or sampling fraction) is calculated thus:

K = Population size

Sample size

After random selecting, the starting point between 1 and

8, every 8 t h name on the list is taken until the 100 names are

complete. If the starting point is 7, the other sample elements

would be l5,23,3 1,39,49 and so on in the list.

3.1.2 Cluster Random Sampling Technique

Cluster random sampling is a process in which the

population is divided into sub-set called (cluster) which

contain two or mare of the population elements. A sample is

randomly selected from among the cluster and the members or

elements therein are studied. In a one-sage cluster sampling,

all the members of the selected cluster are studied. A sample

is taken from the members or elements of the selected clusters

or groups for study in a two-stage cluster sampling. This is

simply sampling from a sample.

3.2 NON-PROBABILITY SAMPLING TECHNIQUES

None- probability samples are taken without giving all

the elements equal or fair chances of being selected. The

criteria for being included in the sample depend mainly on the

judgment of the researcher. Non-probability sampling include

Convenience Sampling

Volunteers Sampling

Judgment Sampling

Quota Sampling.

3.2.1 Convenience Sampling

Convenience sampling entails selecting members or

elements from a population to constitute a sample, based on b

the convenience of the researcher, without any serious regard

for their representativeness. The researcher may choose any

one he considered convenience for him, his friends, the most

accessible places, the most co-operative individuals etc.

3.2.2 Volunteer Sampling

A researcher can take his sample from among those

elements in the population that willingly offer- themselves

from the application of his research instrument. The

cooperative attitude of such volunteer makes it easy for the

researcher to collect information from them.

3.2.3 Judgment Sampling Technique

In judgment sampling, the researcher selects elements

from the population on the basis of his presumption that they

are suitable for the study or representative of its population.

3.2.4 Quote Sampling Technique

Quota Sampling is one in which the elements are

arbitr&ily picked in correspondence to the known proportions

of the different group into which the population has been sub-

dividgd .

QUESTIONNAIRE TECHNIQUES

1) Open-ended questions

2) Multiple choice questions

3) Dichotomous questions

4) Scaling questions

Open-Ended Questions Techniques: Open-ended questions

techniques are constructed in such a way that the respondent

is free to express his answers in own vocabulary as he thinks

appropriate to the questions. The question do not suggest any

alternative answers from which the respondent concloose of

for how long you been long in this cly?

Multiple Choice Questions: Multiple-choice questions

present, either in the question themselves on immediately

following a list of alternative answers from which respondents

are to choose.

Dichotomous Questions Technique: Dichotomous questions

present in such a way that the respondents can only answer

in either of two way- yes- no, did didn't agree-disagree etc. b

(iv) Scaling Questions: Scaling questions which are

designed with increasing or decreasing magnitude type of

response alternative build into them, for the purpose of

measuring respondents attitude or feelings to issues. Which

are usually emotional in nature. Examples are as follows".

How do you see the exam

Very good, good, poor, bad, very bad.

INTERVIEW TECHNIQUES

Structured interview

Unstructured interviews interviewed

Depth Interview

Focused interview

(a) Structured Interview

This type of technique has a standardized (formal) list of

questions, which will be posed to every respondent in the

same order.

(b) Unstructured Interview

Though there exists a list of questions to guide the

interview here, he is allowed to throw in some follow up b

questions on his own, so as to proble.

(c) Depth Interview: Here, the interviews is conducted in a

free style using a combination of semi-structured and

unstructured questions to encourage the respondent to

discuss the subject matter rater then answering questions.

Focused Interviews

In this type of interview the respondent is skillfully

helped to recall and concentrate on a particular experiences or

incidence.

Observation Techniques

Controlled observation

Uncontrolled

Participant

Non-participant

controlled Observation

A controlled observation techniques is one in which the b

things to be watched are closely defined.

Uncontrolled

The investigator watches whatever he feels is relevant in

an, uncontrolled observation. He is neither given instructions

on the specific aspect of the things he is to observe nor a

format for recording them.

Participant Observation Techniques

In a participant observation, the investigator becomes a

part of the situation he is to watch. This enables him to gain

an intimate and in depth knowledge of the situation.

Non-participant Observation

The investigator does not apply any role are get involved

in the situation he is to watch his aloofness may head to

disinterest and rob him of some vital information which only

participants can obtain. However, he is more able to record his b

observations then when he is a participant.

CHAPTER FOUR

CONCLUSION

There are many ways in which data can be collected. The

researcher designs instrument for a specific research, which

could form a kind of primary source of information. Data

gathering is the process or procedure designed for generating

and collecting of data required for problem solving.

The statistical tools to be used for date gathering are

called the techniques. The instrument to be developed will be

based on original information. This, it must have to be

subjected to rigorous test such as validity/reliability.

Therefore, data techniques are any device constructed for

recording or generating of data. It is the means for generating

and collecting the pertinent information to be used for solving

the research problem. Adequate and reliable data collected

contribute a lot to knowledge or solution of life problems. A

researcher should arm himself with adequate courage,

patience native wisdom in order to achieve a better result.

The

secondary

researcher is interested in disobliges of data

and primary data.

The primary data are the field data and are more difficult

to collect then the secondary data, which are essentially

documented data.

REFERENCES

IYIOGWE S. 0 . [Mackins], Awoke M.U. [2000]; Research Methodology, A practical approach to project report writing for Nigeria Universities, Nigeria: Folsun technologies publishing company.

OLAKUNLE K. OLAKUNORI [2000]; Successful Research [theory and Practice], Enugu: Computer Edge Publishers

UCHE MODUM [1995]; Management Information System [Analysis and Design], Enugu" Fourth Dimension publishing Co. Itd.

ONWURA .E. ANEKE [1998]; Research Methods; Enugu: Gostak printing and publishing Co. Itd.

INTERNET SERVICES; Data gathering techniques for science and research techniques.

PANEL SURVEY A panel is a group of individuals who have agreed

to provide information to a researcher over a period of time (Tull and Hawkins: 1976:297). This means that data can be collected two or more times. The data collected at the different time periods can be analyse in order to measure trends and changes in the opinions and attitudes of the respondents with respect to a particular issue.

There are continuous and interval panels. In a continuous panel, the members are made to. report specified behaviours or! a regular basis. Behaviours that , can be measured in this way include people's television viewing patterns, media habits, and purchase patterns. Audiometer can be fixed to people's television set to automatically monitor their preferred channels and programmes. A diary can be provided for the respondents to record their nxdia habits and purchase patterns. Whilc the audiometer is an automatic rescardl instrument, the diary is a self -administered it~strument, resembling a questionnaire. In an interval panel, the members agree to respond to stated number of questionnaire sets during the life span of the panel.

DISADVANTAGES OF PANEL SURVEY 1. The response rate is high, even when mail is used

as the means of contact. 2. Sampling V does not need to be undertaken each

time information is to be collected.

3. Trends and forecast of behaviour can easily be done.

DISAVANTAGES OF PANEL SURVEY 1. The panel members may not be representative of

the population. 2. Some respondents demand gratification before

agreeing to serve in a panel. 3. The = ... . .!. anqlysis , . . of the information collected from

p&el.&_isp~en tedious and costly. F;s :-:. ?- : l>

> . .- ,

ique of watching or quence and details of a researcher wants to e after the occurrence

erson's dining table servation is the only

rrnation. A good example .* i s .+. the ; information on young children's food or

.! p$mbyei : ~ t ;,oth+rtimes, observation is used to - supafernent - elf .preferred to other means of data collekon. An example is when information is needed on how employees do their work or socialize at work place.

For observation to be used as a means of data collection, the basic conditions which must be satisfied are as follows:

1. The behaviour to be observed must be such which occurrence is predictable, repetitive or frequent.

2. The behaviour should be possible to observe. 3. The time taken for the manifestation of the

behaviour must be short.

TYPES OF OBSERVATIONAL APPROACHES Observation can be carried out in any of the

following ways: 1. Controlled, 2. Uncontrolled, 3.

Participant, and 4. Non-participant

1. CONTROLLED OBSERVATION A controlled observation is one in which the

things to be watched are closely define. Such an observation can be structure or use mechanical devices. In a structured observation, the investigator has a pre- knowledge of the aspects of behaviour or situation he is to watch. A printed form, known as check-list, can be prepared to help guard against the investigation's focusing on irrelevant issues. A typical check-list contains a number of items or behavioural aspects of which the investigator is expected to indicate presence or absence after watching or checking. Investigator's bias is highly minimized in structured observations.

~echanical observation makes use of such devices as video tapes (which record actions, scenes, and sounds, all at the same time). Audiometer (which

records the time radio and television sets are tuned on, and the stations tuned to ), eye camera (which records the movements of the eye), pupilometer (which measures the changes in the size of the pupils of people's eyes), and psychical manometer (which measures minute emotional reactions in people, through changes in their rate of perspiration), among others. Mechanical devices make the recording of observation more effjcient.

2. UNCONTROLLEDIUNSTRUCTURED OBSERVATION The investigator watches whatever he feels is

relevant in an uncontrolled observation. He is neither given instructions on the specific aspects of the things he is to observe nor a strict format for recording them. He could therefore be carried away by his personal interests or forget to record some of his observations.

b 3 . PARTICIPANT OBSERVATION In participant observation the investigator

becomes a part of the situation he is to watch. This enables him to gain an intimate and in-depth knowledge of the situation. He can become carried away by his own role or be biased in what he focuses on or records. Participant observation can be controlled or uncontrolled.

(d) Determination of the sample size (e) Sampling procedure/sampling Design /sampling

plan ( Administration of Research Instrument

SCOPE OF STUDY AND AREA OF COVERAGE The purpose her is to define the jurisdiction of the

study in terms of time, demographically and geographical elements, having in mind the research problem and the associated variables earlier stated.

SOURCES OF DATA Two sources of data are generally used" for

research. These are the secondary (documented) and primary (fresh or original) sources. The places, people or things from which each of these two major sources of research data are obtained must be described under this section.

POPULATION OF THE STUDY Population in research statistics can be described

as the entire number of people, objects events and things that all have one or more characteristics of interest to a study. It is the target of the study for collection of data. The population of a study can be large or small. Before he begins to collect information, the researcher must clearly define the population for the study. To be complete and useful, the definition of a study's population must specifj its elements (who or what

general constitute it. The forms in which they e'xist or can be found for contact purposes.

If, for example, the study is titled: An app,aisal 1 of the senior Academic Staff Development Programme of the Ebonyi State University, Abakaliki. The population can be defined as all senior academic staff presently under the employment of Ebonyi State University, Abakaliki.

A correct definition of the population of study will heip us for comparison of research findings, with those of similar findings.

DETERMINATION OF THE SAMPLE SIZE The motive behind sampling is to use the

information obtained from a part of the population to take decision on the whole census. Obtaining information from all the members or elements of a population is generally more accurate than sampling. But it is costlier, more time consuming, and sometimes impossible. Because of these, most researches put reliance on sample, rather that census.

One of the major problems of the researcher is deciding on the appropriate size of sample for his study. The larger the sample size, the more accurate the information to be derived from it, and the higher the costs of obtaining the information.

The size of the sample for any study must therefore consider the desire accuracy of the sought information and available resources (in terns of finance and time). Though many methods are available for

determining sample size,. none of them can be considered as purely scientific or acceptable to all research practitioners.

- n = Desired sample size N= Total Population E - acceptable error limit (#0 05)

Another statistical formula which can be used in place of the methods above is that of Churchill ( 1976: 302-3 14) which says

N = the sample size to be determined Z = the assumed !eve1 of signific~nce or

confidence level. X = amount of variation to be allowec! in the

estimated sample mean. PSD = estimate population standsrd deviation.

Lets assume that, Ebonyi State investment company limited, has the following population of senior staff 1 18, hte-mediate staff= 328 acdJxicx s%ff= G?, total no. of staff = 1135. Determine the sample size, using Yaro Yamani formula.

SOLrnION According to Yaro Yamani (1964 : 280) to

determine a sample from a population n = N

1+ (Ne2)

Therefore, if one is studying the above organisation with a population of 1135,296 people is the sample that he 1 study, which represents the population.

Another formula is the one stated below as Size = sample size

Population size

SAMPLING PROCEDURE The researcher should specifl the methods by

which each of the units will be selected for measurement or collection of information. Two basic methods of sampling are probability and non-probability sampling. Tne researcher should state now the chosen method was actuaiiy employed in the study r'or the benefit of the inquisitive readers.

METHOD OF DATA ANALYSIS Data analysis is normally riot done in chapter

three. Methods used in analyzing the data collected in the study is done here. It r r ? q be dcscriynti~e st~tistlm or inferential statistics, atimes both.

Descriptive statistics are generally used for summarising the data being presented, while inferential statistics are used for the estimation of population values, evaluation of the relationship between two or more variables and testing hypotheses.

For the presentation and description. of the data, tables, percentage, bar charts, pie-charts, histograms, frequency polygons, means (averages), modes, medians, and variances can be used.

Inferential statistics tools include the chi-square (X2), t-test, F-test, regression analysis, and correlation analysis, among others. The researcher should describe how they were actually used in the study.

FUNCTIONS OF RESEARCH DESIGN IN A RESEARCH WORK

According to Black and champion (1976) a research design performs the following functions.

(a) It provides the researcher with a blue print for studying research problems.

(b) It dictates boundaries of research activity and enable the investigator to channel his energies in specific direction.

(c) It enables the investigator to anticipate potential problems in the implementation of the study.

(d) It helps to provide some estim~te of the cost of the research.

In concluding our discussion on research design, it is important to observe that a research is easily executed when it is guided by a design, defined simply as a plan that specifies how data should be collected and analysed (Black and Champion, 1976)

SAMPLING TECHNIQUES The two basic alternatives are probability and

non-probability sampling techniques.

PROBABILITY SAMPLING

The variants of probability samples include: (a) Stratified random sampling (b) Simple random sampling (c) Systematic sampling (d) Cluster sampling.

A probability sample is one in which the selection of the sampling units is based on chance, and for which the chance of each unit being selected is known. Probability sampling makes it possible for a sampling error to be calculated. This merit of allowing for the situation which require generalization to be made about

- . . , -. .

97

Judgment Samplinfz or Purposive Sampling: In judgement sampling, the researcher selects elements from the population on the basis of his presumption that they are suitable for the study or representative of its population. The efficiency of this method of sampling is much dependent on the expertise of the person making the selection. According to (Black and Champion, 1975). As observed earlier, purpose sampling is very useful in certain research situation when the goal of generalising on the population is not needed.

Volunteer Sample: In this method, the researcher may request people to volunteer and can only base the study on the volunteers.

'4 researcher can take his sample from among those element in the population that willingly offer 1 themselves for the application of his research instrument. The co-operative attitude of such volunteers makes it easy for the researcher to collect information from the population whose general behaviour pattern is different from that of introverts. This makes the findings of studies using information obtained from volunteers biased and unacceptable to some critics.

Convenience Sampling: This entails selecting members or elements from a population to constitute a sample, based on the convenience of the researcher, without any serious regard to their representativeness.

CHAPTER TEN

COLLECTION OF DATA

The collection of data is a very vital and exciting activity in the research process. After a researcher has got a researchable topic, done his proposal and selected

- -

his sample, the next problem be faces is how to go about the collection of data required for his topic of study. It is important to remark that data are not collected for collection sake. Each of them bears a direct relationship to the hypotheses, which also have relationship with the problem under study.

The coliection of data is the greatest of the effort of the researcher. If things go wrong here, the data collected may become unreliable, they would go a long way in making the fihal research report a worthy contribution to knowledge or the solution to life fioblerns. And to avoid hindrance in the collection of data, the 'researcher should arm himself with adequate courage, patience, native wisdom, and a good research design when collecting data.

The researcher is interested in two types of data, primary and secondary data. The method of data c~llection to be employed in any study depends to large extent on the research design objective - descriptive, causal or explorative and to a small extent on the availability of funds. Data which are not related to the study should not be soliected no matter how interesting or how easily available they may be.

- - * L - - --e cp&on& s h o X c o n x n an introductory

background which should contain respondents personal data, which include age. qualification, ranWposition, department or discipline etc. Efforts should be made to provide alternative answer to questions like age, religion etc. e.g. What is you age ----------- 10 - 15 years, 15 - 20 years, 20 -25 years and above.

STEPS IN DESIGNING QUESTIONS IN A QUESTIONNAIRE.

No procedure exists for automatically generating a good questionnaire. Much of what is found in literature on the subject of preparing a good questi0nnaire.i~ based on the experience gained in the field.

(a) The researcher should bear in mind the presentation and analysis of the information to be collected whilt dsking the questions.

(b) The questions should be asked bearing in mind the research problem. This is best done by focusing the question content on the research hypotheses or research questions.

(c) The questions asked should elicit specific answers. Questions which use such teems as: normally", "sometimes" frequently, and often are indefinite, ambiguous and difficult to answer and analyse.

(d) When every pasonai information is sought, it should be done towards the end of the

questionnaire when adequate report has been established with the respondents. The words and sentences to be used in the questions should be easy to understand. They should not have double meanings or implied alternatives not meant by the researcher. As much as possible, questions with suggested alternatives answers should be employed. These stimulate the memory of respondents and are easy to analyse. Whenever multiple choice questions are asked, the last alternative answer should be "others" please specify, if the suggested alternativ, are non exhaustive. This accommodates those respondents whose alternative answers are not included among the suggested ones. The questionnaire should be well typed and reproduced on good paper. It shculd normally not exceed two or three pages, so that it can be completed within ten minutes by the average respondents. The constructed questionnaire should be pre- tested before being used to collect data. The pretest (dr pilot study) with a small number of the respondents will help the researcher to know the suitability of the questions or response format

used.

TYPES OF QUESTIONS

109

- --

METHODS OF DATA COLLECTION

These are two broad methods of data collection available to a researchers, primary method of data collection and secondary sources of data collection.

PIUiV1ARY DATA: Primary data are called field data. All data collected for a specific purpose by the researcher from the field are known as primary data. The most important methods of collecting primary data are questionnaires, observation, interview, telephone and panel surveys.

QUESTIONNAIRE SURVEY A questionnaire is a list of questions or

statements, which require respondents to make a reply. Besides, a questionnaire may simple be defined as a data gathering instrument in which respondents are given standard or uniformed questions.

The replies may be recorded by a field worker (or the researcher) who administers the questionnaire or the respondents himself. With the aid of a questionnaire, the research can obtain information from respondents with or without personal contact. Where the sponsor of the research is disclosed by the field worker who distributes it, the questionnaire is called an undisguised questionnaire or survey.

A disguised questionnaire, on the other hand, is one in which the identify of the sponsor of the research

is not disclosed. The disclosure of the identity of the sponsor of a study may sometimes result in respondents bias. This often delays responses.

Questionnaires can be administered directly to respondents or indirectly through mails (mail questionnaire) where mail is to be used (mail questionnaire or mail survey), the questionnaire is posted to the respondents who are expected to return it after their response through the same means mail questionnaire reaches more widely scattered people faster at relatively low cost. But its return rate is lower than when field assistants are used.

The designing, distribution and collection of I questionnaire is generally cheaper than other means of

data collection when per contact costs are considered. , The analysis of data collected through questionnaire is I

1 I, also usually relatively cheap.

STEPS IN QUESTIONNAIRE PREPARATION

1. The address of the researcher should be written on the top right side of the questionnaire.

2. The purpose for which the research is been carried out should be stated in the questionnaire.

3. A guarantee of security of the respondents should be included in the introductory letter part of the questionnaire.

4. Economics of words and space when structuring questionnaire.

(b) The bias of field assistant who record the answers is reduced.

(c) The suggested alternative make it easy for respondents to answer these questions.

DISADVANTAGES (a) Respondents who have more information than the

suggested alternatives are frustrated. (b) The order of presentation of the alternatives

answers can introduce response bias. (d) It is difficult to collect information of omitted

alternatives, if all the possible alternatives are not listed.

DICOHOTOMOUS QUESTION In a dichotomous question, the choice is limited to

only two answers e.g. "Do you feel that EBSU is doing well in the training of students? Yes ( ) No ( )

Dichotomous questions present questions in such a way that the respondents can only answer in either of two ways, Yes or No, did or did not agree (or) disagree, and male or female.

Example: 1. Have you ever been to Ezillo Town? 2. What is your sex?

rnVANTAGES 1. They are easy to contract and answer 2. They are easy to tabulate and analyse.

DISADVANTAGES 1. ' Some respondents may not be fully decided on

either of the two alternatives.

INTERVIEW SURVEY

Interview is a data gathering instrument that enable a seeker of information to have an in-depth knowledge of an issue of concern, through a face to face interaction (a times by telep5one conversation) with the provider of such information. Put differently, interview can be regarded as a method of data gathering technique refers to a face to face communication which a researcher asks respondents questions to obtain answers relevant to the research problem.

Beside, interviews survey is the calling on respondents in order to collect information from them through conversation. The investigator or field assistant who makes the call, asks the questions and records the answer is called the interviewer, while the respondents is called the interviews, The interviewer can call on the respondents at home, at the place of work, in the street, or anywhere. The list of the question together with the spaces for recording the respondents answer is called interview. schedule. The interview guide is the list of points or topic which are to be covered by the interview. Specia!ized training is often required of the interviewer prior to his being sent to the field. This will ensure that

There are available to the researcher our major types of question or response formats. These are:-

1. Open-ended questions 2. Multiple-choice questions 3. Dichotomous questions - 4. Scaling questions

OPEN-ENDED QUESTIONS Open-ended questions are constructed in such a

way that the respondent is free to express his ariswer in own vocabulary, as he thinks appropriate to the' questions. Open-end questions is a form of unstructured question, the responses vary as respondents vary.

The question do not suggest any alternative answers from which the respondent can choose. This is why some researchers refer to this type of questions as free-responses or free answer questions. Examples are:

Why did you buy your present bicycle? What is your highest educational qualification? For how long have you been living in this city?

ADVANTAGES

(a) The respondent is not influenced by the researcher in the choice of words to use in answering the questions.

(b) Questions can be answered in considerable depth.

(c) The respondent is spared the potential frustration of being confined to pre-test set of response.

DISADVANTAGES

(a) It is difficult to subject the question to statistical analysis, since its quantification may be difficult.

(b) Ths respondents level of education and expression can greatly affect their answers.

(c) Investigators bias can be introduced into the responses when recording or reviewing them.

MULTIPLE-CHOICE QUESTIONS

Any question that provides three or more responses options, can be called a multiple choice question. In multiple choice, the response are many, and all should be, valid options unlike the West African Examination Council (WAEC) questions that only one is correct. All questions in the multiple choice question goes "How would you rate the level of performance of Ebonyi State Agricultural Development programme" Low ( ) High ( ) very high ( ) very Low ( ). Another example = What is your marital status?

(a) Married (b) Single (c) Divorced.

ADVANTAGES (a) The answers are easy and cheap to tabulate.

1. The behaviour to be observed must be such which occurrence is predictable, repetitive or frequent.

2. The behaviour should be possible to observe. 3. The time taken for the manifestation of the

- behaviour must be short.

TYPES OF OBSERVATIONAL APPROACHES Observation can be carried out in any of the

following ways: 1. Controlled, 2. Uncontrolled, 3.

Participant, and 4. Non-participant

1. CONTROLLED OBSERVATION A controlled observation is one in which the

things to be watched are closely define. Such an observation can be structure or use mechanical devices. In a structured observation, the investigator has a pre- knowledge of the aspects of behaviour or situation he is to watch. A printed form, known as check-list, can be prepared to help guard against the investigation's focusing on irrelevant issues. A typical check-list contains a number of items or behavioural aspects of which the investigator is expected to indicate presence or absence after watching or checking. Investigator's bias is highly minimized in structured observations.

~echanical observation makes use of such devices as video tapes (which record actions, scenes, and sounds, all at the same time). Audiometer (which

records the time radio and television sets are tuned on, and the stations tuned to ), eye camera (which records the movements of the eye), pupilometer (which measures the changes in the size of the pupils of people's eyes), and psychical manometer (which measures minute emotional reactions in people, through changes in their rate of perspiration), among others. Mechanical devices make the recording of observation more eff?cient.

2. UNCONTROLLEDIUNSTRUCTURED OBSERVATION The invstigator watches whatever he feels is

relevant in an uncontrolled observation. He is neither given instructions on the specific aspects of the things he is to observe nor a strict format for recording them. He could therefore be carried away by his personal interests or forget to record some of his observations.

*3. PARTICIPANT OBSERVATION In participant observation the investigator

becomes a part of the situation he is to watch. This enables him to gain an intimate and in-depth knowledge ofthe situation. He can become carried away by his own role or be biased in what he focuses on or records. Participant observation can be controlled or uncontrolled.

PANEL SURVEY A panel is a group of individuals who have agreed

to provide information to a researcher over a period of time (Tull and Hawkins: 1976:297). This means thzt data can be collected two or more times. The data collected at the different time periods can be analyse in order to measure trends and changes in the opinions and attitudes of the respondents with respect to a particular issue.

There are continuous and interval panels. In a continuous panel, the members are made to. report specified behaviours on a regular basis. Behaviours that . can be measured in this way include people's television viewing patterns, media habits, and purcha~e patterns. Audiometer can be fixed to people's television set to automatically monitor their preferred channels and programmes. A diary can be provided for the '

respondents to rccord their media habits and purchase patterns. Wllilc the audiometer is an automatic rcscclrch instrument, the diary is a self -ndministcred instrument, resembling a questionnaire. In an interval pancl, t l~c members agree to respond to stated numbcr of questionnaire sets during the life span of the pancl.

DISADVANTAGES OF PANEL SUR'VEY 1. The response rate is high, even when mail is used

as the means of contact. 2. Sainpling does not need to be undertaken each

time information is to be collected.


DISAVANTAGES OF PANEL SURVEY 1. The panel members may not be representative of

the population. 2. Some respondents demand gratification before

agreeing to serve in a panel. 3. The. analysis of the information collected from

pabeis.ispften tedious and costly. p. .w*- ; '

ique of watching or g the sequence and details of

a researcher wants to e after the occurrence person's dining table servation is the only

" logical 'means of collecting information. A good example is the information on young children's food or

.' t o $ : ~ ~ n ~ . ~ , : ~ t dother times, observation is used to &,kJ-: ,

supphnent - -- prefehed to other means of data collection. An example is when information is needed on how employees do their work or socialize at work place.

For observation to be used as a means of data collection, the basic conditions which must be satisfied are as follows:

or incidence. The interviewer is expected to be armed with appropriate pre-selected questions which will enable the respondent to freely and completely express himself on the focused issue.

ADVANTAGES OF INTERVIEW SURVEY

1. Interviews are generally a flexible means of data collection.

2. Difficult or mis-understcod questions can be explained or re-phrased.

3. It makes proving possible. 4. Additional information on the personal

characteristics of respondents can be &llected. 5. The men?ory of the respondent can be stin~ulated

to aid recall.

DISAVANTAGES OF INTERVIEW SURVEY

1. Locating the respondent and booking and interview with him may be difficult.

2 . Conducting interviews is much time-consuming and costly.

3. The interviewer may be biased. 4. Unstructured interviews are difficult to analyse.

TELEPHONE SURVEY Telephone survey is a type of icterview conducted

through the means of telephone. It has the main features

of personal interviews. Telephone survey can be used to supplement or aid questionnaire survey. The telephone directory can be used as a sampling frame to obtain the list of potential interviewers.

The advantages and disadvantages of telephone survey are similar to those of personal interview survey, with the following additions:

ADVANTAGES 1. It is cheaper than personal interview survey. 2. It is faster than other methods of data collection 3. It can be used at any time of the day. 4. Otherwise difficult or inaccessible respondents

can be contacted through the telephone. 5. The anonymity of respondents can be more

maintained than in personal interviews.

DISADVANTAGES 1. It is limited to only those who have working

. telephones. 2. Respondents are not given adequate time to think

properly before answering the questions. 3. It does not provide opportunity for observing

people's manners while they are answering questions.

he properly understands people's temperaments, conducts himself well before the respondents, and correctly records the answer.

GUIDELINES FOR THE INTERVIEW SCHEDULE

Issues under discussion should be narrowed down to permit the maximization of depth rather than breadth. This is because the depth of issues covered in an interview constitutes one of its merits over a questionnaire. It may not be necessary to seek for respondents* personal data. The interview questions should be logically arranged to capture the major objectives of the research. The interviewer should have mastery of interview equipment, like tape recorder, and other recording materials. Interviewer needs good public relations skills to be able to persuade prospective interviewees, to grant an interview, to be covered by the interview. Specialized training is often required of the interviewer prior to his being sent to the field. This will ensure that he properly understands people's temperament, conducts himself well before the respondents, and correctly records the answers of information supplied.

TYPES OF PERSONAL INTERVIEWS

The four major types of personal interviews are: 1. STRUCTURE INTERVIEW: This normally has

a standardized, formal list of questions, which will be posed to every respondent in the same order. The interviewer is not expected to deviates from this format.

2. UNSTRUCTUED INTERVIEW: Though there exists a list of questions to guide the interviewer here, he is allowed to throw in some follow up questions on his own, so as you prove the respondent more indepthly.

3. DEPTH INTERVIEW: Here, the interview is conducted in a free style using a combination of semi-structured and unstructured questions to encourage the respondent to discuss the subject matter, rather than answering questions.

Considerable skill is required on the part of the interviewer so as to ensure that he completes the interview successfdly without blasing the direction of the discussion (proctor and stone: 1978, P. 88)

4. FOCUSED INTERVIEW: In this type of interview, the respondent is skillhlly helped to recall and concentrate on a particular experience

Source: Field Survey, 1999.

Table 11-5 Radio Station Preferences of Respondents.

Station

RBC

Rad-Co

FM ~ a d i o Five

BBC

FM3

Others

Total

No. of Respondents

548

398

1052

25 1

153

98

2,500

1

Percentage Degree

22 221100 x 360 = 79

16 16/100 x 360 = 58

42 421100x360 = 151

4

10 101100 x 360 = 36

6 61100x360 =22

4 41100x360 = 14

100 360

The data above can be presented pictorially, with the aid of a pie chart, as show below.

Figure 11-1 Pie chart depicting the Radio Station preferences of respondents.

Source: Field Survey, 1999

(ii) Bar Chart Let us assume that the following table was also obtained from the data generated for the "FM Radio Five" study.

Table 11.6 the preference of Ekulu Residents for FM Radio Five from 199 1-1 999

Year No; of People

1991 8010

1992 9520

1993 12610

1994 16800

1995 15005

1996 22560


below.

25000

20000

15000

10000

5000

.

The data in the frequency table above can be presented pictorially with the aid of a bar chart, as can be seen

Figure 11T2 Bar chart showing the preference of Ekulu Residents form FM Radio Five from 199 1 - 1996


When drawing a pie chart, the researcher must ensure that the measurement of the degrees to be used in apportioning spaces in the circle is accurately done. Otherwise, a wrong impression will be given (pictorially) as to the sizes or proportions of the component of the variable measured. Making use of a larbe circle and a good protractor can greatly help to

Most city dwellers are both at home and awake between the hours of six and eight in the morning. Most radio stations in the country air major news and early morning programme at that period. In order to know the extent to which the respondents to this study give attention to radio programmes during this vital radio air period (6.00 am - 8.00am), we asked them to indicate how often they listened to radio during the period in the week. Their responses show that 50% of them do listen to the radio between 6.00am and 8.00am at least five days of the week. Table 7-3 below gives the details of - their responses. .

Table 11.3 Rate at which respondents Listen to radio Between 6.00am and 8.00pm Every Week.

Responses Every Day Five to six days Three to four days Once to twice

Source: field Survey, 1999

No. of Respondents 40 1 850 1037

Less than once Nil Total

Table 1 1.3 above shows that 40 1 (16%) of the respondents listen to radio between 6.00am and 8.00am

Percentage 16.00 34.00 4 1 SO0

173 7.00 13 26 2500

0.50 1 .OO 100.00

every day, 850 of them (34%) listen to it five to six days of the week. Those who listen three to four days are 1037 or 41.50% 173 (7%) and 13 (0.50%) of the respondents listen once or twice or less than once a week

Those who do not listen to radio at all, week in, week out, are only 26 or 1% of the respondents. The pattern of response above shows that the respondents to the study are very much radio conscious between the hours of six and eight in the morning.

Both government and private radio stations exist in the country. These compare with one another locally and with the foreign stations which can be received in Enugu. While six, local radio stations can be received well by an average-sized radio set in Enugu about ten foreign others can be received well.

The radio station that a person will tune to at any particular time depends on which one he prefers in terns of the pro&unme being aired, clearness of reception, and the credibility of that stations, among other things. Early morning news is one of the major programmes that radio stations air and which attract a large a listening audience. When we pressed the respondents in this study to state how many times they listened to FM Radio Fives "6.30 News" in the past one weak (151h to 21" Septeinber, 1999). We discovered that even though only 101 (4%) of them listened to it daily throughout the seven days of that week, a great many of them actually did more than half of the period. This can be seen in table 7-4 below.

*:

Table 1.1-4 Rate at which Respondents tuned to FM Radio Five's "6.30 News" in one week (15' -2lSt September, 1999).

Responses No. of Respondents Percentage Nil 14 0.50 Once or twice 669 27.00

36.00 Five, to six times 32.00 Seven times I don't know 0.50 Total 1 2500 1 100.00 I


Table 11-4 above reveals that 14 (0.50% of the residents did not tune to FM radio Five's 6.30 News during the week being studied remember whether they tuned to the programme or not. Of those respondents who tuned to it, 669 (27%) did once or twice, 900 (36%) did three to four times, 802 (32%) did five to six times, and 101 (4%) did seven times (everyday of the week). '

Enquiries made during the personal interview held with some official of the National electric power Authority -

and some of the respondents revealed that there was no power cut or electric- fauk in the areas covered by the study during the period. This means that those who did not tune to FM Radio Fiver's 6.30 News might have tuned to other stations or none at all.

140

USING PIE AND BAR CHARTS TO PRESENT DATA

Bar and pie charts can be used to present any of the data in section m 11.3.1. above. But both of them are usually used to supplement the tabular presentation of data. This is because of the fact that tables enable a more comprehensive presentation and understanding of data than charts. But chart are easier to read. While bar charts have the special advantage of enabling readers to quickly read and interpret the trends in data, pie charts are very usefbl in showing the proportions of the component parts of a unit or variable. Both are pectoral and easy to construct, once the frequency table fiom which they are derived is produced. Below are examples of how to use both charts to present data.

(1) Pie Chart

Let us assume that the table below was derived from the data obtained from the "FM Radio Five" study.

figure. This should follow the discussion or explanation o f the contents of the table. Charts or figures should, as much as possible, not be used in view of tables.

6. Where the researcher has obtained some additional information through other sources, apart from the one he is presenting, this should be used for providing more classification on the issue concerned during the discussion that follows the display of the related table.

USING TABLES TO PRESENT DATA Tables enable the researcher . to convert

frequencies into percentages, which are then used for comparison. Presenting data with tables is very exciting. I,

Suppose a researcher conducting - a study on the popularity of a particular radio station asked the following questions (among others) in his questionnaire. i 1. What is your highest educational attainment?

(a) Less than first school leaving certificate (b) First school leaving certificate (c) Attended secondary school (d) WASCIGCEISSCE (e) ONDINCEPOST SECONDARY

DIPLOMA (f) HND/BSC/B.ED and above

2. How often do you listen to radio between 6.00 am and 8 .OO am? (a) Every day of the week

(b) Five to six days weekly (c) There to four days weekly- (d) Once or twice weekly (e) Less than once a week ( f ) Not at all.

About how many times did you tune to FM. Radio five's "6.30 an News" within the past one week? (a) Nil (b) Once or twice (c) Three to four times (d) Five to six times (e) Seven times ( f ) I don't know

Assuming that the respondents answered all the questions and that 2,348 questionnaires were returned out of the 2,500 that were distributed, the researcher can proceed to present the data with the aid of tables as shown below.

This study was designed to find out how popular FM Radio Five is among people in Enugu

metropolis. Out of the 2500 questionnaires randomly distributed to the responding residents in different parts of the metropolis, <a total of 2,348 were returned. The response rate ii-shciwn in Table 1 1 =- 1 below. .

Table 11 - 1 Return Rate of Questionnaire

No. of Respondents Questionnaire

Returned 2,348

Not Returned 152

I

Total (distributed) I 2,500

Source: Field survey, 1999

According to Table 1 1.1 above, 2, 348 (94%) of the respondent returned the questionnaires sent to them, while the rest (152 or 6%) failed to return theirs. The high response rate of 94% achieved in the study is due to the two or three repeat calls made on respondents and the possible high interest they have in the study.

In this study, we soughtr to know the hightest educational attainments of the respondents. Their respones show that majority of them have at least WACIGCE O,level/SSCE. The details of their responses are contained in table 11-2 below.

Table 1 1.2 The Education Attainments of Respondents

I Educational Attainments I No.of I Percentage I

13.00 15.00 2.00

1 - ~

WASC/GCE/O'leveVSCE OND/NCE/Post Sc.

1 Total 1 2,500 I 100.00 I

Less than FSLC FSLC Attended Secondary School

Diploma HND/BSC/Bed and above .


3 26 3 73 5 1 988 548

In Table 11.2 above, it can be seen that 326 or . 13% of the respondents have less that the First School

Leaving Certificate, 373 (15%) of them possess the F S L ~ as their highest academic qualification. 51 (2%) attended but 'did not complete secondary school. Those who possess WASC, SSCE OR GCE (O'Level) among the respondents are 988 (39.50%) who have OND, NCE or post secondary diplomas and 214 others (8.505) who have HND BSC, Bed, or higher as their highest educ~tional attainment. From the fact that 70% of the respondents have WASCISSCE or GCE )'level or higher educational qualifications one can simply say that they are highly literate and should be interested in the electronic rhedia.

39.50 22.00

2 14 8.50

and Analysis of Data) in the final research report. It precedes the testing of hypotheses or rigorous analysis of the research questions (where research hypotheses) had been used, the presentation of data is normally elaborate. This is because elaborate analysis is required in such a case. The most important aim of the organization of data, as discussed in the last section, is to get the data well-prepared for presentation.

Very often, the data presentation sectio%r of;' student' research projects if so poorly arranged, readers '

are immediately put off by it. Data presentation is more than showing boring figures or response frequencies to readers. It should explain to them in an exciting way the state of things, vis-a-vis the data collected for the study. Below is a procedure that can make the presentation of research data purposeful, organized, informative and exciting.

1. The presentation ,should be subdivided into parts. This can be done according to the research variables, issues raised in the study, research hypotheses, research questions, subdivisions in the questionnaire, and sets of questionnaire used, among others.

2. The presentation should begin with the introductory data. These are data that relate to the response rate for the study (such as the return rate of questionnaire) and the classification of respondents according . to sex, age, income, educational qualifications, occupation, and status in office, among

others. The purpose of the categorization is to show readers how representative the sample is or how unbiased the study is. This assures readers of the seriousness of the study.

3, As much as possible the presentation of data should be done according to the responses to each question asked or each issue requiring answers.

4. In presenting the date on responses to any question or issue, the researcher should first of all introduce the matter or issue to be dealt with in one or two short sentences. The introduction should be followed by a statement of the investigation done or the question asked (in reported form), and a summary of the pattern of response receive. After these, reference should be made to the table which contains the full data on the responses or issue. These should take about three sentences. After displaying the table showing the response frequencies and corresponding percentages the researcher should proceed to read-off or discuss its ,

contents fully. This pattern should be maintained for each question item or issue until all the data for the study are prescrired . However, care should be taken not to be repetitive in the words used for introducing the questions or matters dealt with.

5. Sometimes in order to achieve more effect or provide more explanation, the researcher may want to

summarize the data contained in the table in a chart or

accurate. Non-response items can be treated as having "don't knows". They can be obtained through telephone or another call on the respondents, inferred dr estimated from other answers, or discarded as a complete non- response. Illegible and ambiguous data are deciphered by the editor or referred to the field assistant or respondents concerned for clarification. The accuracy of the supplied data can be improved upon by checking the internal consistency of the response.

3. Coding Coding entails categorizing the responses qr

answers obtained from the field and assigning data (frequency values) to them. Coding is done in order to get the data prepared for tabulation. It is essential part of editing.

4. Tabulation 'Tabulation involves transferring data from the

coding sheet, test score sheet, or pre-coded questionnaires or interview forms and putting them in tables. By arranging information.' according to their attributes, tables help to facilitate an easy understanding of the relationships among the variables being studied. A good table is that which is relatively simple. To be simple and easy to understand, a table must concentrate on a few ideas, attributes, or variables. Every table must be numbered and titled. The title is always placed at the top of the table.

5 . Creation of New Variables After tabulating and carefully studying the data,

, an experienced researcher or analyst can generate new variables from them For example, the respondents in a firm might have supplied. information of their length of service, the positions.

By combining the responses to these three issue the analyst might come out with (new) information on the rate at which employees are promoted to managerial positions or the number of the firms that hired managers and the ones promoted from the ranks. The use of computers can greatly facilitate the derivation or generation of new variables during the analysis of data.

6. Calculation of Summarizing Statistics. After tabulating the data for each variable and

showing the frequency distributions, the stage is set for the calculation of summarizing statistics. Summarizing statistics are of two major types-measures of dispersion (standard deviation, variance, and range). These statistics or values enables the researcher or analyst to have a better understanding of the data array. The values can sometimes be used in place of the data from which they were calculated.

PRESENTATION OF DATA The presentation of data is the formal display of

the data collected in a research, couple with a brief explanation on the pattern of response to each variable. It is usually the first part of chapter four (Presentation

Two statistical tools are employed for analyzing research data.

These are known as descriptive and inferential statistics.

The analysis of research data is a tedious but interesting activity. It requires painstakingly going through the mass of collected data, in order to extract meanings that would help to solve the research problem. The product of the exercise depends on the quantity and quality of the data collected and the careful employment of correctly-chosen statistical tools for analysis. The old saying "garbage in, garbage out" - comes to mind here. In a situation where the wrong type of data was collected and the sample is not representative of the-populations, wonders should not be expected from the analysis of such data, in terms of generating results that can be generalised or used for prediction. Poorly done analysis can lead to the loss of some information that could otherwise have been generated. Additionally, the resultant "bad information" generated by such a poor analysis can lead to a bad decision. Good decisions are based on correct data and sound analysis.

THE REDUCTION/ORGANISATION OF DATA Since it is the analysis that will bring out the real

meaning of the collected research data, effects must be made right from the beginning of the study to ensure that ' 1

. .I the process of analysis is not made difficult. . ( ' ~ ' -1

The pre-testing of the research instrument will greatly assist the researcher here. Once the data are secured, the researcher will need to first of all, sort and get them arranged in such a way that it will be easy to evaluate them and observe (or test) the relationships existing among the variables highlighted. This is what data reduction or organization is all about. Simply put, dw- reduction is the process of getting the fresh Ailected research data ready for analysis and the calculation of summarizing or descriptive statistics (Tull and Hawkins, 1976:477-488). The steps involved in data reduction are: (1) Field controls, (2) Editing, (3) Coding, (4) Tabulatiion, ( 5 ) Creation of new variables, and (5) Calculation of summarizing variables.

1. Field Controls Field control are those devices which are aimed at

ensuring that field assistants measure all the laments of the sample correctly and properly record all their identification characteristics, such as name, address and bf contact, in a sample books. This will enable the editor or researcher to be able to make necessary verifications of the information supplied or the additional information from the respondents if the need later arises.

2. Editing : Editing entails going through the turned-in

' research instrument (such as questionnaire) in order to ensure that the data requested are supplied, readable, and

129

overloaded. Case , .

improving and

. . , . . .

. .

3. They are .less time cons,um methods of data collection^.-^.^:^^

4. Respondent bias is greatly

. -; $

1. The concentration on only one or a few situat~ons .

makes generalisation of findings very risky. . 2. The accuracy of the data collected and findings is

reduced by the non-random - : selection , of samples and the subjective measureme'nt employed.

CHAPTER ELEVEN

DATA ANALYSIS

THE MEANING AND IMPORTANCE OF DATA ANALYSIS

The analysis of data is the breaking down of recorded research information into meaningful parts for critical examination, extraction of otherwise hidden information and making of inferences about the relationships existing among the parts. The otherwise hidden information or meanings which are obtained from analysed data enable the researcher to draw conclusions for the study. Based on these conclusion, recommendations are made to guide the decisions of the sponsor of the study or those who are interested in its results.

Data do not become meaningful or useful until they have been analysed. It would not be possible to answer the research questions or verifj the truism of the research hypotheses until the collected data for the study have been scrutinized, using systematic and verifiable techniques. These techniques are statistical in nature. Statistics, as a body of knowledge, is concerned with the collection, presentation, critical evaluation, and interpretation of numerical data. It enables a researcher to draw general conclusions from data. It enables a researcher to draw general conclusions from data and make predications about what will happen when certain

popular Iabc education ant

rratory types 'of experimen~ conauucu ,111

d business researches. -5 m-

2 FIELD EXPERIMENTS Field experiments are carried out in the natural

setting of the things studied. This ensures a high degree of realism. The control of the ' independent and extraneous variables is often difficult in field experiments. The investigator may fail to secure the co- operation of subjects (such as retailers who may be expected to stock a particular brand of products and sell at a specified price). Such extraneous factors as strikes, weather conditions, political and social disorders and changes in fiscal and monetary policies cannot be controlled by the experimenter. Yet the variables can greatly affect the results of a field experiment. One of the most popular field experiments is test marketing, which is often conducted prior to the comercia1isation of a newly developed product.

3. . COMPUTER SIMULATION Simulation is the manipulation of the variables in

a model so as to generate information or learn something about the real system being portrayed by the model. A model is the hypothetical specification of a set of variables and their interrelationships designed to represent some real system or process, partially or hlly Kotler, 19980 : 118). Simulation is increasingly being

carried out through computers because of the growing complexity in the environment. Through the use of computers, a great number of variables can be manipulated at the same time, with quick results, computer simulation has not yet attained much ,,

popularity in the social and management sciences and education. But it is very popular in the pure and quantitative sciences, such as theoretical physics, chemistry, mathematics, statistics, and computer science, where it is used for "desk" researches.

ADVANTAGES OF EXPERIMENTATION I. When properly used, it is the most fruithl and

reliable method of data collection. 2. It has the least respondent bias of all the methods

of data collection.

DISADVANTAGES OF EXPERIMENTATION I . It is very rigorous to design and execute. 2. It is usually very costly to conduct.

CASE ANALYSES The case approach refers to the intensive analysis

of a single or few situations of a phenomenon for the purpose of obtaining information that would lead to a greater understanding of that phenomenon and the possibility of generalisation. The intensive analysis done in a case study enables the researcher to discover variables and relationships that would otherwise be

!I6 SYSTEM DOCUMENTATION

(e) Listen more thali talk: yet say enough to keep thc discussion going along the right lines. Avoid leading questions, aim towards short questions and long answers. i t ' i t is felt necessary, prepare beforehand draft of the questions. Any suspicion of misunderstanding calls for a rephrasing of the question or posing it again latu.

(f) Interview only one person at a time, thus eliminating arguments between staff, and reducing the diffidence of the more reserved persons. If a manager calls in members of his staff to answer specific questions, encourage tlxm to depart as soon as this has been done. Similarly, discourage a manager from attending an interview involving one of his staff as this tends to inhibit the latter in his rep1 ies.

(g) Control the- interview by minimizing digressions, separating opinions from facts, and not aliowing generalizations to obscure the true situation.

(h) Do not attempt to cover too much ground in m e interview. After one hcur, at the most, temporarily conclude the discussion and .arrange to resume in the near futcre. Long interviews suggest that either the interviewee is being allowed to digress or that the discussion is becoming bogged-down.

(i) Conclude with a brief resums 3f the ground covered, asking if any major points have been omitted, arid leave an opening for further

TOOLS FOR FACT-FINDING & ANALYSIS 117

discussion. Any dubious information is verified by sending a memorandum later; with operational staff this is best done via the departmental manager.

Observation

Observation should be the best way of obtaining facts about what actually happen because workers could be saying one thing during interview but in practice a different thing is done. But unfortunately the use of observation for fact finding has a ma.jor drawback, which -is workers in the system may consciously select n nat they want to do as long as the observation lasts. Observing entails watching the departmental staff carry out their various tasks in the normal way. But as a rule, people do not take kindly to being observed at their work, therefore their output might not be factual. Also, it is time consuming on the part of

,, the analyst to sit to observe others. Therefore the analyst must have a definite purpose. He should pay attention to interruptions to the normal flow of work, balance of the work:oad, usage of files and documents and informal communication of information. Other f ~ c t finding tools include sampling and measuring.

Sampling

This is a process whereby a sample is drawn from the entire group - usually known zis the population because the population is too large to be measured completely. This sample so drawn is measured in the hope that the outcome is representative of the population. In dealing with sampling theory what the analyst should monitor is the

114 Y YSTEM DOCUMENTATlON

analyst interviews the right people in the right order and with the right questions. Success in an interview is as dependent upon the preparations made for it, as on the way in which it is conducted. Therefore adequate preparations must be made before embarking on an interview. Areas of preparation are

Self preparation

The analyst must ensure that he knows enough about the interviewee's work and responsibilities, and role in the organization to ask the right question at the right level. Awareness of the character and the temperament of the interviewee helps a great deal.

Another essential factor is the creation of a check list - check lists for the information he seeks.

Notifying the Interviewee

It is very important that the in'terviewee is informed about the interview long before the actual date, and how long the interview may last. It is equally necessary that the interviewee knows the purpose of the interview and what exactly the topics to be discussed are, to enable him think about it.

Suitable Environment for the Conduct of the Interview

It is-4ht: analyst's duty to make sure that the interview is as relaxed and comfortable as possible, preferably in familiar grounds but without interruptions or distractions.

TOOLS FOR FACT-FINDIYG & ASALYSIS I!5

For the performance of .an interview, there ate a number of guidelines for the analyst to follow. The under listed are rules%for conducting an interview as outlined by H.D. Clifton in his book Business Data System. pg. 202.

Rules for conducting an interview

(a) Obtain a manager's consent before interviewing his staff and, as far as possible, work downward through the staff hierarchy.

(b) Arrange the time, place and subject of the interview ' yell beforehand SG that the interviewee is able to make arrangements regarding his work, and to collect together documents ma information related to the subject of the interview. If there is likelihood of continual interruptions or eavesdropping, the interview should be arranged to take place away from the interviewee's normal work place.

(c) The interviewer should brief himseGon the position and general duties of the interviewee; and also, of- course, on the subject of the interview. Vagueness in these respects tends to lessen the interviewee's desire to co-operate.

(d) Put the interviewee at ease. by providing a quiet, interrupting-free environment. If a tape recorder is used, explain why and reassire the interviewee that ik use is not for the purpose of subsequent "inquests, "

TOOLS FOR FACT FINDING AND ANALYSIS

In Chapter 7 we briefly touched on the four most common tools .for fact finding which are Interview, Observation, Reading and Questionnaire. The intention of this chapter is to take a closer look at the tools, with a view to understanding how best to apply them for optimum result.

Questionnaire

This method of fact finding should be least used by the system analyst, because of its inherent constraints. The constraints include the difficulty of producing questions that are intelligible and unambiguous to the reader. Most often the readers fail to respond to questions not because they do

112 SYSTEM DOCUMENTAT13N

not want to but because the questions are vague and coi~fusjng, some times contradictory. At best the questionnaire is used to find a relatively small amount of information from a relatively large number of people c,r events especially where the distance poses some problems.

When it is absoiutely necessary to use questionnaire. the design of questionnaires has to be done well if the exercise is to be worrh doing at all. The questions sliould be put into well defined categories. The questions requir: YesINo answers, the multiple choice questions which offer a range of specific answers of wliicli one or more may be ticked, as appropriate to the question and the circumstances. The inclusion of 'other' in the list of possible answers should be critically evaluated. The designer of a questionnaire must also decide for which questions the answers "not applicable" and "not known" should be stated explicitly because they do not mean the same thing. Otherwise for some questions a blank answer might have either interpretation.

Again in formulating questions for questionnaire the designer should pay attention to the language used. The language should be simple, straight to the point, and free from bias. For instance compare the following two questions "Do yo11 fill in the form as soon as it arrives in the department?" and When do you fill in the form?" It is clear that the first question has bias while the second questio~ is preferred. Also avoid leading questions or questions that are based on opinion. For instance it is better to ask how long did you have to wait before seeing the Manager? Then you specify the ranges for the reader to ;ick th'e appropriate range. Instead of "Did you have to wait long to see the Manager*? The reason is that 'how

!ong tlic wait' is dependent on each perwn's temperament. 10 minutes' wait may be too 101:s ior one person but may be nothing to another. Finally on quesrion formulation, the designer sliould make sure tliar the 'rrght questions' are directed LO 'right respondc~i!~' i n order to ensure that facts sought are got.

Apart from pa! ing attention to tlie questions the layout of a quectionnalre should be attractively designed, paying attention to spacing and tlie questions should be logically arranged a ~ ? d organized rather than jump from topic to topic. The cl~.nrlli~ie for tlie return of the ~lucsti01lIlairc ITlllst bc clearly stated. I t is of use recsivilig rile q~~crtio~!nnire after [llc f ; ,c( ,

Reading

This method is usually the first to be used. Reading covers reading of appropriate journals, procedure manuals,

, company reports both internal reports and published annual reports, staff handbooks and departmental records. Initially reading helps the sysreln analyst to get acquainted with the nature of the system and its methodology. It also provides many details about tlie system and tlie details of statutory regulations which apply to tlie work. But t.he problem with reading as a tool for fact finding is knowing really what to read and when to stop. Usually the amount of documentation and literature companies have are limitless. Therefore efforts should be made to see that only the relevant ones are read.

Interview Interviewing is probably the most valuable method if the

14 4

the convenience of the researcher, without any serious regar for their representativeness. The researcher..may choose friends, the most accessible places, the most co-operativf individuals, and the first set of people he encounters, jusf because they would yield information to hini without mucq ado. This type of sampling is often used in explorative studiej and pretests, when the reduction of time and financial costs iJ a main objective. I

2. Volunteer Sampling

A researcher can take his sample from among thosc elements in the population that willingly offerihemselves fool the application of his research instrument. The co-operativt attitude of such volunteers makes it easy for the researcher tc collect information from them. But volunteers are usuall; extroverts, a segment of the population whose genera behavioural pattern is different from that of introverts. Thi makes the findings of studies tising information obtained fron volunteers biased and unacceptable to some critics.

3. Judgement Sampling

In judgement sampling, the researcher selects element from the population on the basis of his presumption that the are suitable for the study or representative of its populatior The efficiency'of this methbd of sampling is much depender on the expertise of the person niaking the selection.

4. Quota Sampling

Quota sampling is one in which the elements a1 arbitrarily picked in correspondonce to the known proportiol of the different groups into which the population has been sul divided. Thus, , i f ten lecturers are to be picked from

112 THE COLL 4 . ,

145 - .,* . and lightly on-surveys. Ho:vever, cost constrain&ay sometimes influence the choice of a let; efficient method or variant of a method over a more efficient ortt. Hereunder is a description of the most popular methods of data collection.

Survey research can be defined as the systematic gathering of infamation from respondents for the purpose of understanding and/or predicting some aspects of the behaviour of the population of interest (Tu// andAlbaum, 1973:3). Usually, the data needed for solving the research problem in :urveys are collected from only a few people or items (sample) cotsidered to be representative of the entire ; population. Most researches conducted by businessq, institutions, and j students in social and management sciences as well as education are . survqs. Types of survey research include questionnaire, interview,

\ telephon_e_~dlpanel surveys. -... **-. -.--.-.-- K - - --

.. - - - - . - - - - - - -------- --

/ /'

/ A questionnaire is a list of questions or statements which --, require respondents to make a reply. The replies may be recorded by \ i

t a field worker (or the researcher) who administers the questionnaire or 1 the respondent himself. With the aid of a questionnaire, the researcher , 1 can obtain information from respondents with or without personal

contact. The questionnaire and/or field worker who distributes it may or may not disclose the sponsor of the research to the respondents. , Where the sponsor of the research is disclosed, the questionnaire is called an undi i naire- or mmey . -A ' disguise a" q u e s s i r e , o Is one in yhich the identity of the sponsor of the research Is not disclosed. The disclosure'of the IdendtY

- of the sponsor of a study may sometimes result in respondenfs bias. 1 Where this is not expected, the sponsor of the research should be disciosed in the quesdonnaire. Thb often quickens responses.

Where a quesdonnaire is to be administered by dlrect per! I

contact with respondents, the use of trained assistants or fleld Wo

may be required. The training w e n to the assistants will result in less respondent bfas and non-response. Where mail is to be used (mail questionnaire or mail survey), the questionnaire .js posted to the respondents who are expected to return it after their response through the same means. Mail questionnaire reaches more widely scattered p e ~ p l e faster at relatively low cost. But its1 return rate is lower than when field assistants are used.

The designing, distribution and coliection of questionnaire is generally cheaper that other means of data collection when per contact costs are considered. The analysis of data collected through questionnaire is also usually relatively cheap. Hence, the great popularity enjoyed by questionnaire survey. But a good questionnaire is not easy to design. The collection is also not easy, as many respondents loathe "filling forms" or may be biased in their response.

?Ips on Questionnaire Desf gn

Questionnaire design is still much more an art than-a science. No procedure exists for automatically generating a good questionnaire. Much of what is found in literature on the subject of preparing a g ~ o d questionnaire is based on the expefience gained in the field. The

-,-I- ---* --- -. - fBlEw~ng tips can be of great help to re.&-cher in avoiding p i % a

L i@or t .---A- to design - - a goo~questionnaire. . _ _ - _ - . _ -.--- __ .-- I I - - - - -- -

/ l < The researcher should bear in mind the presentation and I analysis of The information to be collected while asking the

questions. I

The questions should be afied bearing im mind the research

* problem. This is best done by focusing the question content on

I the research hypotheses or research questlons. I 8. The questions asked should elicit specific answers. Questions \ whkh use such t e r m as Unormallyl" somethneq" " k q ~ e n t l y " ~ b a n d uoften" are lndeflnlte, ambiguous and dlfAcult to answer

and analyse. -- .? - - _ - - _ -

ensure that the space (pie) is shared according to the scores contained in the frequency table.

The space between the bars and the sizes of the base of each bar in a bar chart must be equal, respectively. The space between the bars depends on the number of bars to be shown. The space can be made small or completely closed, and the bars joined if the bar are many and space is a problem. The heights of the bars must be drawn to scale. Sometimes, the different bars or spaces in the pie chart are coloured or shaded differently in order to facilitate comparison.

When presenting data with tables and charts or figures, the researcher should take not of the following:

1 . Every table, chart or figure must be propekly numbered and titled. The title should be simple, short, and correctly depict the data being presented.

2. While the titles of tables are written at the top, those of charts, graphs, figures or plates are written below or at the bottom. The tables and charts in this chapter are good illustrations.

3. The sources of the data being presented with tables and figures must always be given. Tables and f iyres generated from primary data may simply be called "field survey", "research data", or "field data".

A single table should not be made to depict too many ideas or issues at the same time. Complex table should be avoided, as much as possible. It is often an otherwise complex table.

Charts (most-especially, pie charts) should not contain too many ideas or issues at the same time.

Charts should be used in addition to or in place of tables only when they can presents the data concerned better. Othenvise, they lead to duplicity and wastage of space.

The data presented in tables and charts should always be discussed or explained in the text of the research report. The assumption that the data are there in the table for reader to see by themselves is not usually accepted in formal writings.

A table that can conveniently be contained in a single page should not be divided into two for placement in two pages, just because the space in the page'where the table begins is not enough to contain the whole thing. Rather than divide the table into two, the researcher should take the whole table to the next page.

If no hypotheses had been set for the study and the researcher is only being guided by research questions, his comments on the pattern of responses would be a bit more pungent than what we have in section 11.3.1. This is because he would have to draw his conclusions for the study from the percentage comparison that follows the tabular presentation of data. From here he would proceed to state the major findings of the study, consequent to making recommendations. But if hypotheses had been there to guide the study at the beginning, these would have to be tested with the data collected, after their tabular and, or graphical presentation. '

METHODS OF TESTING RESEARCH HYPOTHESES

The testing of hypotheses is the main focus of the analysis of researcher data. Untested hypotheses remain only as tentative explanations or solutions to a research problem. The testing of the hypotheses enables the researcher to prove their authenticity in the light of available data, consequent upon making generalisations from +he research data. For instance, suppose a research student has collected data from certain farmers on their income from the production of cassava and yam tubers.

The student had earlier hypothesized that the farmers' income from cassava is higher than the one from yam. He would have to test the truism of the hypothesis with the data he has collec$ed from the

farmers, before his earlier claim. (Hypothesis) can be accepted as a statement of fact, as far as the population of the farmers he is studying is concerned.

Various method exist for testing research hypotheses. These methods., tests, or tools are statistical in nature. They can be divided into two broad groups- parametric and non-parametric tests.

1. Parametric Tests Parametric tests are those statistical techniques which assume the sampled population to be of known form. For any analyst or researcher to be able to adopt these techniques for the testing of any research hypothesis, the knowledge of estimation of the means, variance, or some other population parameters is a prerequisite. The following assumptions are also made about the population from which the research sample is drawn. i. The observation (sample units) must be

drawn from a normally distributed population.

ii. The observations must be independent of one another. This simply means that the selection of inclusion of any-unit or element from the pepu!etior, as part of the sample must not bias the chances of the inclusion of any other unit. This calls for the use of

4. Very perscnal questions (such as those relating to respondent.' age, income, number of children, status and name) should not be asked unless they are vital for the study.

.,,-. '. ----.--- . 5. when ;e& personal information is sought, it should be don;)

1 towards the end of the questionnaire when adequate-report has ' kbeen established with the respondents7! In addition, the' ' suggested 3ternative answers the respondents are to choose from should have enough intervals, so-as to make them not to feel too exposed when responding to the questions.

____^C---w.-.-- . . -*.. -.. -n?r y 6. (The words and r e n $ - ~ . c ~ s . ~ ~ b ~ ~ d ~ ~ ~ t h e questions .should&e

!kasYasYtp,undemand They should not havi;double meanings or implied alternatives not meant by the researcher.

- - . - . . - 7. ,As much as possible, questions-aith..suggested-alte~ativel

i i. answers --_ - - -. should .- be ernpl.oy.ed.;These stimulate the memory of respondents and are easy to analyse.

-. - -. . . . . . . - - - - . -.

_- .- .-. - . __ 8; <whenever multiple choice questions are

alternative answer should be "others, please -specify,"-if-the suggested alterijat&c~ .a~e . tmtexhausr id~hl r accommodates those respondents whose alternative answers are not included among the suggested ones.

9. ~ e a d i n ~ or loaded questions should be avoided. For example, such a question as: "Do you have the hygienic habit of bathing with soap every day?" has a tendency of generating a positive answer.

10. The questionnaire should be well-typed and reproduced on good paper. It should normally not exceed two or three pages, .so that tt can be completed withln ten minutes by the average respondent. . .

1 1 . Every questionnaire should go with a forwarding letter, introducing the researcher, the objectivcs of the research, soliciting the co-operation of the respondent, andassuring him of anonymity. This normally quickens response and reduces non-response.

12. The constructed questionnaire should be pre-tested before being used to collect data. The pretest (or pilot study) with a small number of the respondents will help the researcher to know the suitability of the questions and/or response format used. Amendments should be made where necessary before the full distribution of the questionnaire.

- . ______-._.__.,_<-0-.- - ' ' -. - '- - - -- - . .. _ ym of Questions \

There are available to the researcher four major types of, _.__- -- --- 1- __-_._ - - ~ ~ u e s t l a n s or response fy-ma* or a combination of two _-._ or more _-_ ___ of which he maLempl&~while-designing his questionnaire. Khese are ! i) open-ended. (ii) multiple - choice (iii) dickogm-ousLand .--- (iv) ____- sca l ing j ue~tions._.j"T-"-'~' - ' --

L- --

__._ I - - - -- ._^_ i

emended questions are constructed in such a way that the respondent is free to express his answers in own vocabulary, as he thinks appropriate to the questions. The questions do not suggertany'

.- altepacve ansye~&ofiyhich the respondent can chooser This is why some researchers refer to t h ~ s type of questions as free-response or free-answer questions. Examples are:

Why did you buy your present car? What is your highest educational attainment? For how long have you been living in this city?

---------_--. ._ _ b&amtages - -

- 1. The respondent is not influenced by the researcher in the ,

Y choice of words t o use in anyer ing the questions.

149

/'

/ '- Questions can be answered in considerable depth.

/ 3. The respondent is spared the potential frustration of being : confined to pre-stated set of response alternatives. i

I

Disadvantages

, 1 . Investigator bias can be introduced into the responses when recording or reviewing them

2. The respondents' level of education and expression can greatly affect their answers

I 3. Since every respondent's answer is unique, the tabulation and ; analysis of the supplied information become much tinsc-1 :-'

consuming and costly. I .

- , - I - .r 2. - - --- ..

"(ii) Multiple-Choice Questions - .

Multiple-choice questions present, either in the question '

j themselves on immediately following a list of alternative answers from ' which respondents are to choose. Examples are as follows: _ ,- I What is your marital status?

(a) Married (6) Single ( c) Divorced a Does Blue Omo. Blue Surf. Blue Elephant. Jet. or

..._ . Nasco Detergent provide the cleanest wash?

1 I . The suggested alternative make it easy for respondents tb answer these questions. 1 2. The bias of field assistants who record the answers-Is redud.

i

I \ 3. The answers are easy and cheap to tabulate and analyse. - 1

j Disadvantages

I 1 . The construction of the questions takes a considerable effort. 1

2. It is difficult to collect information of omitted alternatives, if all ,

the possible alternatives are not listed.

3. Respondents who have more information than the suggested alternatives are fruhtrated.

i4 . The order of presentation of the alternatives answers can, -- introduce response bias (position bias). - _ . /-

1 ?(iii) ~ i d o r o m o u r Questions

i Dichotomous questions present questions in such a way that the I :espandents c a i only answer in either of two ways - yes - no, did - ;

didn't, agree - disagree, and male - female. This is an extreme form of multiple-choice questions. The following are examples:

Have you ever been to Abuja? What is your sex? Did you attend a co-educational school?

Advanrages . ---

1 . They are easy t o construct and answer 2. They are easy to tabulate and analyse

Disadvantages

1. , Some respondents may not be fully decided on either of the".: two alternative answers. For example the question: "Do you 1 expect to own a house within the next three years?" may eliclt

. .. -- .. " 1 don't knowN, -. . - .. ~-

-2___

llmaybe~lll l l i f~rr I and "Probably" in addition to "yesn and "no", depending on the type of respondents involved.

2. Bias in response may result from stating the questions positively or negatively or from implied alternatives.

(hi) Syling Questions I

Siding questions are those questions which are designed with increasind or decreasing magnitude type of response alternative built into the$,, for the purpose of measuring respondents' attitudes or feelings to, Issues. which are usually emotional in nature. Examples are as follows:

How do your see the recently announced Federal Governmentlsl budget?

very much Some what I am Very Little Not at all Favourable Favourable Indifferent Favourable F-

Favourable 5 4 3 2 1 Unfavourable

Favourable + 2 + I 0 - 1 - 2 Unfavourable - - - - -

or, lising the Likert scales, the respondents can be told t o choose from the following: Agree very strongiy, Agree fairly strongly, Agree, Undecided. Disagree. Disagree fairly strongly, Disagree very strongly, and Don't know.

Advantages

1. It Is easy for.respondents to fill 2. It Is easy t o tabulate and analyse 3. It is better than other formats when measuring attitudes and

feelings. 4. It cautions respondents to be consistent in their answers.

Disadvantages

1 . The alternative responses are usually difficult to construct 2. Some respondents may find it difficult to understand and

choose from among the alternatives.

A questionnaire has three parts. These are: ( 1 ) identification data (respondent's name, status, and address, and field assistant's name, date, and time), (2) sought data (the basic information needed in the study), and (3) classification data (the respondent's age, sex, education, income, and marital status). The purposes of the three parts are to ensure the scrutinization of the respondents, solution to the research problem, and proper reading of meaning into the data generated.

5

Interview survey is the calling on respondents in order to collect j informatfon from them through conversation. The investigator or field assistant who makes the call, asks the questions, and records the answer i Is called the interviewer, while the respondent is called the interviewee.

'

i The interviewer can call on the respondent a t home, a t the place of , b r k , & d w u e e ~ The list of questions togemer 6

spaces for recording the respondent's answer Is called Interview schedule. The lnterview guide is the list of points or toplcs which are

to be covered by the interview. Specialized training is often required of the interviewer prior to his being sent to the field. This will ensure that he properly understands people's temperament, conducts himself well before the respondents, and correctly records the answers or information supplied.

. - -.. Types of Personal 1nter:iews - -_. - --

I -----7 - The four major types of personal interviews are:

Structured Intemiew: This normally has^- a ~tandardized~formal list of questions which will be posed ,to -every *respondent in the same order. The . interviewer , is' not expected ro deviate froin this format.

Unstructured Interview: Though there exists a list of questions to guide the interviewer here, he is allowed to throw _ in some follow up questiocs on his own, so as to probe the respondent more indepthly.

Depth Interview: Here, the interview is conducted in a free style using a combination of semi-structured and unstructured questions to encourage the respondent to discuss the subject matter, rather than answering questions. Considerable skill is required on the part of the interviewer so as to ensure that he completes the interview success full^ without biasing the direction of the discussion (Proctor and Stone:. I 978,p. 88).

Focused 9ntemiew: In this type-of interview, the i respondent is skillfully helped t o all and concentrate on a IY%K \., panicular>per!e-nreo_r.incidence. e interviewer is expected

\ to be armed with appropriate pre-selected questions which will \ enable the respondent to freely and completely express himself \ on the focused issue. 1

Cr--_ -.---_ _ - - - - -_ 154 /------

! Advantages of interview Survey 1 . Interviews are generally a flexible means of data collection. 2. Difficult or rnis-understood questions can be explained or

t re-phrased. I

! 3. It makes probing possible. 4. Additional Information on the personal characteristics of

respondents can be collected. 5. The memory of the respondent can be stimulated to aid recall.

Disadvantages sf Interview Survey

1 . Locating the respondent and booking an interview with him,' may be difficult.

2. Conducting interviews is much time-consuming and costly ' 3. The interviewer may be biased. 1 4. Unstructured interviews are difficult to analyse. 1

- .---.- - .. - - ___------ C. TELEPI40NE SURVEY

Telephone survey is a type of interview conducted through the means of telephone. It has the main features of personal interviews. Telephone survey can be used to supplement or aid questionnaire survey. The.telephone directory can be used as a sampling frame to obtain the list of potential interviewees.

The advantages and disadvantages of telephone survey are similar to those of personal interview survey, with the following additions:

Adwaaatages 1. Yt is cheaper than personal interview survey. 2. It is faster than other methods of data collection. 3. It can be used at any time of the day. 4. OtheMlise difficult 01- inaccessible respondents can be contacted

through the telephone. 5. The anonymity of respondents can be more maintained than in.

personal htervlews.

1 . It is limited to only those who have working telephones. 2. Respondents are not given adequate time to think properly

before answering the questions. 3. It does not provide opportunity for observing people's manners

while they are answering questions.

A panel is a group of individuals who have agreed to provide information to a researcher over a period of time (Tull and H a w h : /976 :297). This means that data can be collected two or more times. The data collected at the different time periods can be analysed in

, order to measure trends and changes in the opinions and attitudes of the respondents with respect to a particular issue.

There are continuous and interval panels. In a continuous panel, the members are made to report specified behaviours on a regular basis. Behaviours that can be measured in this way include people's television viewing patterns, media habits, and purchase patterns. Audimeters can be fixed t o people's television sets to automatically monitor their preferred channels and programmes. A diary can be provided for the respondents to record their media habits and purchase patterns. While the audiometer is an automatic research instrument, the diary is a self-administered instrument, resembling a questionnaire. In an interval panel, the members agree to respond to a s ed number of questionnaire sets during the life span of the panel. . '"5 Advantages of Panel Survey

1 . The response rate is high, even when mail is used as the means of contact.

2. Sampling does not need to be undertaken each time Information is to be collected.


Disadvantages of Panel Survey

1 . The panel members may not be representative of the population.

2. Some respondents d ~ m z n d gratification before agreelng to serve in a panel.

3. The analysis of the information collected from panels is often tedious and costly.

/ .---- -"'.---- '--.----- I---_ -- .---.. _--_ ---__

:,:- 6.2.2 QBSERVATION RESEARCH ----..__

i

Observation is the technique of watching or monitoring and recording the sequence and details of incidents of particular sorts that !

1 a researcher wants to study. Observation can take place after the [ occurrence of the incident, such as studying a person's dining table or I

; bedroom after leaving it. Observation can also take place as the : 1 incident is takin'g place, such as a shopper's purchase activities in a I supermarket. S o m e t i ~ ~ o b s e r v a t i o n is t h e - only logical means of

colle~!$_informatiou. A go63 example is the information on young children's food or toy prefxences. At other times, observation is used to supplement or preferred to other means of data collection. An example Is when information is needed on how employees d o their work o r socialise at-wo_r,k-plage- ..--- _ ------.- ---- - - F6r observation to be used as a means of data collection, the basic conditions which must be satisfied are as follows: ---\

\ \

I * ) : 1, The behaviovr to be observed must be ru& w h t h &currence 1 i Is predictable, repetitive o r frequent. i ' 2. The behavtour should be possible to observe i

i 3. The time taken for the manlfestatkm of (h behwfour must be , i short. i

157 .- . , - '6 ,',,A 1 --- c' - -

Types of Observational Approaches - - - - - I 1

\

i Observation can be carried out in any of the following ways: 1 ( 1 ) controlled, (2) uncontrolled, (3) participant, and (4) non- ' ~k.rti:ipant. . ,._ _ - .

_ __-----.- . - . .. . --;. -.----..=--I--

. _----- - ---- -.-.* 1. Controlled Observation

A controlled observation is one in which the things to ;

be watched are closely defined. Such an observation can be '

structured or use mechanical devices. In a structured observation, the investigator has a o L b e haviour .or_- known i s chek-list, can be prepared to help guard against the investigation's focusing on irrelevant issues. A typical chek-list contains a number of items or behavioural aspects of whkh the investigator is expected to indicate presence or absence after watching or checking. Investigatorsf bias is highly minimized in structured observations.

Mechanical observation makes use of such devices as video tapes (which record actions, scenes, and sounds, ail at the same time), audiometer (which records the time radio and television sets are tuned on, and the stations tuned to), eye camera (which records the movements of the eye), pupiflometer (which measures the changes in the size of the pupils of people's eyes), and psychogalvanometer (which measures minute emotional reactions in people, through changes in their rate of perspiration), among others. Mechanical devices make the recording of observations more efficient. b

- -- ...__ Uncontrolled/Unstructured Observation 7,

The investigator watches whatever he feels is'relevant

1- - - - I ! , in an, uncontrolled observation. He Iz neither given instructions , ___I/

ings he is to observe nor a strict could therefore be carried away

by his persbnal interests or forget to record some of his observations.

-------.------ - -- r--- 3. Participant Observation

in participant observation, the investigator becomes a part of the situation he is to watch. This enables him t

and indepth knowledge of the situatJ$f?c:; -.

carried away by his own role or be biased in what he focuses on or records. Participant observation .can be controlled or uncontrolled.

_ -______..._&. - - - - .

Non-participant Obserwation

The investigator does not play any role or get involved in the situation he is to watch. His aloofness may lead to disinterest and rob him of some vital information which only participants can obtain. However, he is more

-ZI than when he is a

observation can also be c o n t r o ~ n c r o i i e a . _.-I---- .__. -- ---.-- .. -

Advantages of Observation Research i -

I 1 . It Is usually a more accurate measurement.than a survey where both of them are applicable.

2. It can be used to supplement other methods of data collection. 3. It pi-ovides more current Information than other methods, with

the exception of experimentation.

\ '*

Where it is well-designed, observation method can greatly eiiniinate respondent and investigator blas,

. - . - .-

Disa&tantiqps of Observation Weseaaci;

1 . It fs very expensive in terms cf time an( ! finance. 2. The meihod is only limited to things that can be observed, :

repetitive, and have short occurrence duration. 3. . The interests of the investigator can influence what he sees ..

-- - and/or records. . - . - . - - - - - - -

- . - 6 3 EXPERIMENTATION

r

,J- - Experimentation entails the cjeliberate manipulation of one or : more conditions (variables) by an investjgator iil such a way that its impact upon one or more other conditionskan be measured.

I j Experimentation is a cc.mrnc;n pilenomenon in life. Common 'i i 1 experiments in our social lives include dating, tests, trying on of shoes \ I and clothes in boutiques,' and test-marketing of products. An \ I i experiment differs from an ordinary observation in that' the . 1

\ ; experimenter br investigator can make changes in the situation so as to . \,

? ,<.-- see .-. what .-. - - effect ------___-.-----L.- each change has (Rachnan andMescon: /985:-589J+-! I' L Two -..-- groups .--. ... --.--- a r e _ U ~ ~ ~ ~ . . u ~ d e ~ - a 5 s e _ E , t ~ ~ ~ , ~ h o u g h both of them a n similar in all conditions, one group is exposed to a specific condition (variable), whiie the other (c~ntrolled group) is not. This enables the measurement o f c x m i relationships among the manipulated or changeable conditions. There are three types of experiment4 environments. These ;re laboratory, field, and simulated environments.

Laboratory experiments are carried out under, conironed operational conditions. These aliow the manipulation ' of one or more of the independent variable; without much (If \

-effects of Ystm Laboratdry experiments

experiments in both time and financid resources. Performance tests (existing in the forms of aptitude and achievement tests are a popular laboratory type of err.erirnent conducted in education and business researches. .

~. --. - .- - _ _ _ _ __,> , _ - .- ..*. . . -. //-- - ' . --.

2. , Field Expeuhcazs

Field e3periments are carried out in the natural setting of the things being studied. This ensures a high degree of

i realism. The control of the independent and -- r-.,-, . . extraneous, - -

[\id'.t&s_fi ofte~&fficqI_t-i~fieed - __ experimenw- -__.-- The mvestiga tor may fail to secure the co-operation of subjects (such as retailers who may be expected to stock a particular brand of products and sell at a specified price). Such extraneous-factors as strikes, . weather conditions, political and social disorders and changes in fiscal and monetary policies cannot be controlled by the experimenter. Yet the variables can greatly affect the results of a field experiment. One of the most popular field experiments is test marketing, which is often conducted prior to the commercialisation of a newly developed product.

- - - -7- . Computer Sirnaiation \ \ \ Simulation is the manipulation of the variables in a \ model so as to generate information or learn something about \ the real system being portrayed by the model. A model is the / \ hypothetical specification of a set of variables and their

interrelationships designed to reDy-esent _s_om_e- real- syge_m or i \.g_gess,* partially-~.r_.fdIyGtIer, -. __ -1 980: 6 la ~ i m u l a t 6 6 s

increasingly being carried-out -thrati@Th-cGbuters because of the growing complexity in the environment. Through the use of computers, a great number of variables can be manip~iated at the szme time, with quick results, Computer simulation has not yet attained much pop~larity in the social and managemint'-

' %

sciences and education. But it is very popular in the pure and quantitative sciences, such as theoretical physics, chemistry, mathematic's, statissics, and computer science, where it is used fof "desk" researches.

I . When properly used, it is the most h i t fu l and reliable method , of data collection. I

2. It has the least respondent bias of all the methods of data ,' collection.

>" .*_.----- //-- - - .

isad advantages of Experimentation . 1 1

1. It is very rigorous to design and execute. 2. It is usually very costly to conduct - . . . . . . . - - - - . - - ~ - - .- - ---* ,---._- 1;

- - -.-. - - .=--___ '--T--~ -.---

,:6.2.4 CASE ANALYSES *-. .- --"7

I i

The case approach refers to the intensive analysis,of a single or j ,; few situations of a phenomenon for the purpose of obtaining j [ informati& that would lead t'o a greater understanding of that / \ phenomenon and the possibility of generalisation. The intensive analysis I

done in a case study enables the researcher to d i ~ c ~ e j j a r i a b l e t - a m i~.relationships.. that . ~ ~ u l f i & h ~ ~ i s e h e - o v & & ~ Case analyses can

help a researcher in improving and simplifying the hypotheses to be used in a wider study. This explains why most applications of case dnalyses are In pilot and explorative studies.

-.------ ----- -.-. ----. _ _-- 162 .* --- ----_____ -.

A&ant&es of Case Analyses 7,

1 ' ' They produce grpat lnslghts into hypotheses when used) in pilot studies. / 2. They are cheap to conduct. : L

3. They are less time consuming than the o the~ methods of data '

collection. , , 4. Respondent bias is greatly mlnlmized.

I

~ i ~ a n t a g e s of Case Anawes

1. The concentration on only one or a few situations makes generalisation of findings very risky.

, 2. The accuracy of the data coilected and findings is reduced by. the non-random selectlon of samples and the subjectJve~ measurement-process employed.

Appendix A Areas under the normal curve

02 .03 0 1 .05 06 -----

Appendix B

DATA GATHERING TECHNIQUES FOR SCIENCE AND RESEARCH TECHN~KY S... Page 2 61.12

* 7 1 he reasoning and argumentation in empirical science and research is generally on raw data and structures, calculatio~ls etc. derived f;om raw data. In the present contri- we shall try to present dirferent tecliniques to gather raw data. Anyway, all the interestinq attributes of an empirical word cannot be dcscribed or ~neasured in seconds and c e n t i n ~ e t v ~ r Somc phcnomena or thc rcal word shall be described using some more sophisticated fbrtwjht structures then numbers, sometimes we shall describe the situation using many items, summarized into one figure. Not all operations with tlle~e formal items are invariantrela&d to t1,e abandolment upon the concrete formal mpresenta~ion of attributes and tlmefore nncf crU results of calculation are meaningful, l'hereibre the measurement they and scaling p r o b l w . pcrform the essential role in the data utilization.

On the first part of this contribution we shall try to present different techniques -ti, gather raw data. In the second part we shall indicate the main problems connected with measurement and scaling theory. '

All the content of this col~tribution is rocused to the conceptual-analytical r e s e a r u theory-testing researches, qualitative, interpretative and new theories creating researches 4. an to artefact coilstructive researches according the clksification in JBrvinen (1999). Doec 4

. .4*

cover the mathematical researchei, mentioned in his monograph, because of the differea+ character of the nlathenlatical reasoning and the special position of mathematical and 10&~~~crl~1 . disciplines betweell 'the scirnces.

b

' .(. 1. Different tecliniques

The most typical data gathering techniques are observation, interview, c lues t io \~ .d~l~ and written material. In addition, introspection video and voice recording, surveillance b3 a coinputer program, pilot tests or prototypes and arteraits can be mentioned. To yse a cdmbinatio!l of various data gathering teclrniques is generally r&ornmended.

'

Observation Everyone observes his environment. Anyone observes the actions of others. Sc icdaA~

er.k relinhle and nhirxtive nhservatinns frnm whirh r.an dr3w valirl inferences Thev t r eq t 44,)

- DATA GATHERING TECHNIQUES FOR SCIENCE AND RESEAKCJ-I TECHNlKY S... Page 3 of 12

---.- --- - - J - - - - . - --"-A .-.-.,-A- .--..- .,-.---A ---. ---.. .--a- ...-----.-- ". -.-- - J ----

observation as a part of measurenlent procedure. Observations must be naturalistic, obsefl'krs . must be immersed in ongoing realistic and natural situations and must observe behaviour a L t - occurs in the raw. Direct observation may be more reliable than what people say in many instances. It can be particularly useful to discover whether people do what they say they d- , LV

behave in the way thy claim to behave. - The following checkiist (see Bell (1993), p. 119 - for the observation of actions 4 .

human behaviour only) is recominendable to follow: - Decjde exactly what you need to know (list of all topicslaspect is recommendablk).

~oiis ider why you need this informatiun (examine your list and remove any itcm that is d- dircctly associated with the task). - Is observation thc best way to obtain the information you need (consider altcmatives). .

Decide which aspects you need to investigate (process, product, interactions, . . . ?) Rcqucst permission (official channels, individuals concerned) - -

Devise a suitablc grid, checklist or chart (consult and adapt where necessary a published examples)

'

Consider what you will do with the data and whether the data contain information, c o n b l d d - (Is it likely to produce anything of interest? Will the data be the sufficiently complete to enable you to come to any conclusion) Pilot your technique and revise i f i~ecessar~ . Prepare carefully before observation (Draw the plan, indicate an arrangements and layou6 prepare the etalons for measurement, if required ) -

.' Select a suitable place and period of observation. ~ei&ember that no grid, howcver sophisticated, will tell the full story (Try to place the. e \ I d - in this organizational context. Analyse and interpret data, consider what the facts indicate or imply.

-

1.1' 111 terview '

Interview can be described as a conversation between interviewer and responded w-.

, the purpose of eliciting certain information from the respondent. Type 01 interview diffes depending on the research approach used. In the theory testing approach completely - formalised interview, where interviewer behaves as much like a machine is recomrnenda 1,G . For the theory creating approach completely informal interview may be more suitable in *r/

to get and present as all-round views as possible. The mle of interview in the constructive - approach differs depending on whether we have a building of new artefact or evaluation caw-

In tile first case the discussion between the developer and user is essential, in the second c q L .J' -

P a w 3 -

DATA GATHERING TECHNIQUES FOR SCIENCE AND RESEARCH TECHNIKY S... Page 4 of 12

if criteria to be used in evaluation are predetermined, the fornlalised interview can be appropriate, but if such a criteria are still sought.for, the informal interview is recommendable.

In thc fornl&sed or struc~ured interview interviewcr tries to stay as natuml as p m s i k ~ and interact all respondent in.thc similar way. Based on research framework and 11ypotI~ciS the interviewer carefully plans and forn~ulates questions. In the informal interview the interviewer tries to gather description of life-word of the interviewee with respect to intc~prelation of the meaning 01 described phenomena. Often such an interview is "semi- structured", it is neither a free discussion nor a strictly structured questionnaire. The . - establishing an easy relationship by warming up beibre the interviewing session is usefu L L... both types of' interview.

The following main aspects to understand the qualitative research interview cart k ~ - del.i'n~itated:

Concentrating to the interviewee's life-word

Seek to understand the meaning of phenotnena in the interviewee's life-word

Interview shall be .

Qualitative

Descriptive '

, Specific .I'

Presuppositionless

Focused on certain themes

, Open for ambiguities

Open io changes depending on interpersonal interactions.

Bell (1993, p. 99) presents a nice long checklist with 21 points for a good interview. 4

1.2 Questionnaire Questionnaire in the paper or electronic format contains structured or unstructurd

questions, with the intention to be responded by the selected people. Questionnaile is moi+ use data gathering technique in the survey studies, one of the theory testing research ' ,i.

approaches. Questionnaires with open questions can be also used in the theorylcreating studies. The difference between structured and open questions correspond to the similar diSference between structured and informal interview and almost all instructions given f ~ . / interview can be applicable also for questionnaires. .

DATA GATHERING TECHNIQUES FOR SCIENCE AND RESEARCH E C H N I K Y S... Page 5 of 12

. ; The practice for standardized for questionnaires in survey research is based on t k following assun~ptions:

- 1. The meaning of the questions shall be shared by a majority of respondents

2. The respondents mostly understand the sti~nulus 6r phenomenon under investigativll -

roughly equivalent way

3. 'lke responses will be given in a manner, which allows the researcher to interpret d compare them. -

Page 4

Some recommendable criteria for question writing applicable for interview too follows: Is the question related to the research problem or research objectives? All the items of t h ~ schedule should have some research problem relationship. -

Is the type of questibn appropriate? If the information needed can be obtain also by observation, the more objective approach shall be selected. Is the question clear and unambiguous? The ambiguous statement or item permits or invh@ -

differing responses resulting from alternative interpretations. .' Is the question not a "leading question", which suggests answer? . .i* - Does the question demand lcnowledge and information that the respondent does not have :7 control the invalidity of response due to a lack of information, it is wise to use information filter questions, i.e. "Whether a respondent k ~ ~ o w s what xyz is and means?" -

Does the question demand personal or delicate material of controversial nature that respondent may resist? In the positive case the special tecllniques 10 obtain such inform&,

- ,which is needed. is the question loaded with social desirability? With respect to sensitive behaviours, peep-@ wi!l tend to under-report performance of social undesirability behaviours and overlreport - social desirable behaviours. When the behaviour of interest is not particularly sensitive, s d -

reports tend to be quite accurate. - The surveys that require users to evaluate or make judgments about information systems 4- their effect on specific work activities can produce misleading results if respondents do uot '

interpret of answer questions in the ways intended by the researcher can be fount!, n I-Iufm,@ - and,Conca (1 994).

DATA GATHERING TECHNIQUES FOR SCIENCE AND RESEARCH TECHNIKY S... Page G of 12

1.3 Written materials and other documellts Such a sources, written, printed on recorded using various multimedia techniquq, c h

bc divided inio primary and secondary sources. Priinary sources are those sources, which

. came into existence in the period under research. Secondary ones are interpretation of ev& of this period, based on primary sources. Secondary sources may be originally collected 1 5 u - r

another purpose and in the process of accomplishing another activity. To locate the possible documents can be sometimes a difficult task, which is o f 0

under rated in present time. The cyberspace including Internet now offers huge possibili& + easy collect text. But the Internet is rather passive media. The agent software can be utili st& '

in data and knowledge gathering. This new area also causes new kinds of ethical issues W rescprcher. The amount of documentary material that you can study will inevitably be influenced by the amount of time that is available for the stage of.research. It is not u s u a ~ i . ~ possible to analyse everything and so the researcher must decide what to select, based on WL research problem.

The analysis of documents can be divided into external and internal criticjsm. E k h d criticism aims tq discover whether a document is both genuine and juthentic. Internal criticism aims to analyse the context of the document.

2. Measurement The measurement in the broad sense can be described as a method of describing -%L

selected segment of a real word using some formal structure, which can be processing usk-7 inathematical operation and/& information technologiqs. Measurement is a powerful t0oL.W .

the objectification and of any research. In the mostly frequented special case the formal object are interpreted as a.numbers (natural numbers, integers or real numbers) of the sel& dodain, so called measures. But in oRe11 not all properties of numbers can be interpreted ui properties, which are n~eaning-full in the selected segment of real-word. This phenomenon '&

Page 5

induced by the abandonment in the miasurement. We can measure the length &he table?.^ cei&netre or by inches, the temperature in Celsius or Fahrenheit degrees and the a c h i e v e d of the student during test using scale 1,2,3,4 or using percentage from 100 to 0. Only sucq results derived by computation from measures (numbers) are meaningful in the real .word, which are independent related to the abandonment of the measurement. The degree of t h e mentioned abandonment can be described using group of so-called admissible trailsforin& of ~&isul-es. ~ h / s moun identifv the scale tvne of measurement. The tvne scale used in

aistinguisned basis 01 the gropp 01 admissible trans~ornlatlon 01 measure values, their -

algebraic propei-ties,and the types of mathematical operations permissible for each of thcq .

.A tzomitlal scale assigns numbers only as an identifying label, to each object or 5& $ objects under study. Assigning letters or different some pictures instead of n u m b s

' 4.

serve just as well. Such measurement is in the fact only a categorisation. The s e t q+ U admissible tl-ansfornlations is the set of all one to one n1appings on thc domain 06 measures. The inequalities between the assigned measures (numbers) and the . arithmetical operations between them have not any empirical meaning and it is

' necessary to avoid from the attempt to interpret such operations in the real word --

An ordinal scale assigns numbers to objects, which are rank-ordered with r e s p e a -h .

some characteristic or our preferences. The classical example of ordinal scale typ t F . the classification of pupils in the school or the structured questionnaire, where responses "strongly agree", "agree", "neutral" disagree" "strongly disagree" are coded by numbers 1,2,3,4,5, respectively (or qualifiers: extremely, quite, s l ightl~, - neither or equally, slightly, quite and extremely, combined with adjective sets is claimed to form such a scale. Generally it is not any argument for the presumptiv~ ak

'that differences in such steps are equal each to others. The group of all admissibk - , transformations is the set of all strictly monotone functions In the ordinal scale t k . .I-

inequality symbol between the measures has an empirical meaning, because it e~pfi-i the preference, but arithmetical operations have not empirical meanings. An itztetval scale is obtained if we can define unit of measurement such, that . a difference of scale values has a empirical mealling regardless of whether it is -

a

represents (say the difference between 10 and 9 is the same-as between 7'and 6 &+d h - 5 etc:). The most known example of interval scale is temperature measurement utL , '

- Celsius degrees, Fahrenheit degrees or Calvins. The set of all admissible transformation consists of the functions'of the form y = a 0 x + b, where a > 0 a arbitrary constants. In the interval scale the arithmetical has the empirical mea~ii-7-

b , but it is not meaningful for example to speak of one scale value being twice or t l v a times as large as another. A ratio scale is in the fact the interval scale having further an absolute zero p o i n t i w 'only one freedom in the measurement in this type of scale is the selectiou of unit[ ' example meters, yards or versts). The set of all admissible transformation consis-fS -% . .I.

the functions of the form = a x, where a > 0 is arbitra~y constant. The ratio s d L preferable to use if it is possible, because almost all arithmetical operations can 'M interpreted in the real word and have an empirical meaning. -

DATA GAI'WEKING TECHNIQUES FOR SClENCE AND RESEARCH TECHNIK ... Page 10 of 12

ur r!reasust.s). 3t.c LY urrrrarrL( I Y 1 0 , . The r a ~ w earl ut. surrkxlrrlt.s rwre r~-wrrarrr~ru~L~c.

Churchill (1 9.79) mentioned three reasons for this: a) Individual item usually have considerable uniqueness or specify in that each item

tends to have only a low correlation with the attribute being measured and tends La

relate to other attributes as well. b) Single item tends to categorize objects (peoples) into relative small number ofgrok(~S . c) Individual items typicaIly have considerable measurement error. Reliability tends

increase and measurement errors decreasks as number of items in a combination .increases.

The author's reco~nrnendation is to add the further important reason. .'

id) If our intension is to map our preferences into the measures of the attribute of a g i J h objects and our prefcrences forms only a partial order not a week order (not comparable objects exists), it is not possible to realizc any ordinal type scale measurcrnent by individual measure. But the multi-item ordinal type scaIe ,

-

. , measurement is possible. The minimal number of necessary items is the topologies-( dimension of the respective partial order. See VaniEek (2000).

Churchill (1 979) suggested the eight steyspt-ocedure for developing multi/item measures Steps 3 and 5 are the data collection steps, and hence we shall not describe here in more detail. Step I ; SpeclJL domain of the cort;l.truct. After the researcher has performed a literature s w &

- and no acceptable measure was found, should have good reason for proposing an a d d i t i w d new measure. The researcher'rnust be exacting indelineating what is included in the definition and what is excluded. For example Bailey ahd Pearson (1983) defined user information satisfaction as the sun1 of feelings or affective responses to distinguidhable iacwg of the computerlbased information product and services that provide within organization* Step 2: Gerlerate sanlple of items. The researcher must generate items, which capture the domain as specified. Those methods that are typically productive in exploratory research WL

generally productive. Churchill (1979) also recommendable to use literature search to fi-d .. ow combined variable has been defined previously and how manydimensions or

i3 A#yiib?a?€ a 06 "'lgyyg $&geq ,j?Faf{iy'ag yg a set of i f e w which tap e ' . of the imenslons of the construct at issue. ,By ~ m c o ~ i j o m t ~ i ~ ~ ~ ~ ~ ~ & ~ ~ r ~ ~ ~ ~ ~ n ~ ~ f icpjie - . . . - . - . . . . - - . .. . . . . . . . . . . . . . - . . . .. . . . - - . . . . . . . . - - . - . , . . - . " . . .. . -

' -.' .> lneaning in statements in the item pool, the researcher provides a better foundation for t h e eventual measure. Near the end of the statement developing stage the focus would shift ttB

item editing.

Page 8

/'4 > DATA GATHERING TECHNIQUES FOR SCIENCE AND RESEARCH TECHNIK ... Page 11 of 1.

. .(. Step 3: Collect data. Step 4: Pzzr~j) rireasures. The purpose of the purifying process is to analyse whether a c e f G -- item belongs to the domain of the construct or not. The split-half test, where items are - randon~ly divided into two groups and correlation betwe'en groups is calculated, also m e t * b the.koodness of the multi-iten1 measure. There is one basic p r o b l e ~ with using i t : ~ i f f e r d - results may be obtained depending on how the items are split half. Crombach's alpha me- '

reliability coefficient, defined by = ( k / ( k - I ) ) I J I ( - ~ O ~ I O ~ ) ,

I

wherc k is number of parts (items) in multi-item measure, * CJ ,' is the variance o f t he*. , i., 3

~4 0 ,' is the total variancc of the multi-item measure is recommendable to use for the selec&& 4- j

possible ways of splitting a set in half. The ijctor analysis can be used to test whether t h e items are located to one factor or not. The items that much deviate from the rest bft11e se t WLL

kxcfi~ded from the whole set of items. Step 5: Collect data. Step 6: Assess reliability of new data. The major source of error within a test or measure 5 S W sampling of items. In the case of sampling is appropriate, and if the items "look right", &.L measure is said to have face or conte~gt validity. Cronbach's alpha is the basic statistic fof determining the reliability of measure based on internal consistency. Step 7: Assess construct validity. Churchill (1979) wrote that the specifying the domain 06- 4 L

construct, generating items that exhaust the domain, and subsequently purifying the resu&2? measure should produce a measure, which is content or face valid and reliable. It may or ha. not produce a measure, which has construct validity. The construct validity is most direct? related to the question of what the instrument is iil fact measuring, what construct, trait of concept underlies a person's berformance or score on a measure. To establish the constrw+ .

validity-of a measure, the researcher must determine the extent to which the measure correlates with other measures designed to measure the same thing and whether the measwe bel~kves as expected. - -

Step 8: Develop nonrzs. A raw score on a measuring instrun~ent niay be particularly not infornlative about the position of a given object on characteristic being measured because 4 .

.inits in which the measuring instrument is expressed may be unfamiliar. The position oi + individual on characteristic can be estimated by comparing the person's score achieved by other peop1e:The technical name for this process is "developing norms". -

On the end of the contribution let us to mentioned that multi-item measurement, -%A - is measurement by vectors of nu~nbers, not only.by individual numbers is not a last step k U .

possible useful abstraction. More sopliisticated and complicated mathematical objects -

structures can be used for the don&n of possible measure values. Such a structures can LL also processing using the co-temporally information technologies. Such a generalized ineasurement and the theory of scale types in this general situation is still open area for theoretical science and research.

quarlrrrative and continuous. With this scaling one can conclude that ran position 3 is twice as strong as rank position 6, for example. We can also denote the position of which the rank is a tenth, say, of a given rank position, Ratio scaling is common in the physical sciences. In attitude tests, common 1 in educational and business researches, interval scaling is used rather thzur ratio scaling because the level of measurement is not sophisticated enough to permit minuscule measurements of infinitesimal changes.

Scaling generally is used to categorise (classify), rank or rate qualitative. data. Scaling instruments (used to scale observations) include the Likert and: Thurstone scales.

2.6.5 Time series and cross-sectional data Another important distinction to be drawn-with reference to the data is

that between time-series and cross-section data. Most data utilized in econometric model estimates are of one of these types.

Time series data measure a particular variable during successive time periods or at different dates. The time period is often a year (i.e. annual data), but it can be a quarter, month, or week (i.e. quarterly, monthly, or weekly data). For other purposes a longer time period is used, such as two years, a decade or longer. Usually the observations are successive and equally spaced in time. The reader can easily recall examples of time-series such as the annual gross domestic product, the monthly bank rate and the weekly security price.

Cross section data measwe a particular variable at a given time for different entities. Just as the "time period" can assume different values in time-series data, the "entity" can assume different identities in cross-section lata. The entities might, for example, be different countries so that criiss- ectional data refer to selected countries. Other cross-section data could refex o situations of such entities as firms, industries, regions, households, or ndividuais at a given date. Such data are frequently obtained from surveys )f the entities involved, e.g. questionnaires mailed to heads of households to )btain data at the household level.

Sometimes cross-section and time-series data are merged or "pooled." . The result could be interpreted as a crosssection of time series or time series! of cross sections. An example is a pooled cross-section time-series data on per capita national income in market prices for selected countries over a number of years. The percapita-income data could be presented in a table in which any column is a cross-section over countries for a particular year, and any row

is a time series for a particular country. Panel data (or longitudinal datu) are a special type of pooled cross-

section-time-serics data in which thc same cross-scction is sampled over time. The data are generally ~ ~ c ~ o ~ R € ~ I pertaining to individual economic agents, such as families or firms. An example is panel data on family expenditure, where certain selected families keep recordsof their expenditures over several years. Most of the data available for econometric research are macrodata, pertaining to aggregates of individual agents. Microdata avoid aggregation problems and allow one to cstimate modcls containing behavioural relations applicable to individual agents. Such data are generally not available, however, since they are expensive to obtain and their publication may reveal proprietary or confidential information.

2.6.6 Experimental versus non experimental data We explained these types of data in 2.6.2. Also, in 2.5.2.1. and 2.5.2.2, we

treated survey research and experimental research methods. A re-reading of the explanations of these two research methods will enable the student to draw a.distinction between non experimental and experimental data Non experimental data are obtained typically from observations of a system not subject to experimental control. By contrast, experimental data are obtained from a controlled experiment - that is, a situation in which the system or process under investigation is isolated from outside influences and, to whatever extent possible, influences on the system are subject to the control of the experimenter. We have already given the examples of an electrical experiment, which is a controlled experiment common in the physical sciences; and an advertisement blitz carried out in one of two similar but separated markets, which is a controlled experiment studied in marketing =d managerial economics.

Typically data utilized in the natural sciences are experimental, resulting from controlled experiments, while in the social sciences they are non experimentd, since the underlying conditions are not subject to control and cannot be replicated. The laboratory natural sciences, including chemistry and physics, typically do utilize controlled experiments. A physicist, for example, performing an experiment in high-energy physics will often utilize an accelerator - a controlled environment providing experimental data that can be replicated. However, physical scientists sometimes make use of non experimental d a t ~ For example, astrophysicists generally cannot perform laboratory experiments but must instead rely upon observations over which

they have nu control. The same is true, in general, of meteorologis geologists, and classical (but not molecular) biologists.

Conversely, while social scientists - including sociologists and political scientists as well as economists - generally rely on observational, no , experimental data (such as those obtained in sample surveys), much as th astrophysicists do, some social science experimtmts have generate experimental data. For example in a negative tax experiment, individuals and 1 families are enrolled in a programme to test the effect of direct subsidies tcr those with incomes below a defined poverty level. The experimenters musb select sites where they can achieve some isolation of the subjects from extraneous influences, and can use control groups. Other such large-scale: social experiments can be conducted to test the influence of health insurance - where it exists - for the poor and near poor and to test the influence of- housing allowances. The experimental data obtained in ti~ese experiments typically relate to groups of individuals or families enrolled in specific programmes, and they generally include observations over several years. Thus they yield pooled cross-section-time-series microdata.

2.6.7 Accuracy of economic data Sample data are generally used to estimate quantitative population characteristics such as measures of central tendency and of dispersion. Estimators of population parameters are called biased if the mean of the estimates as the sample size increases does not become closer to the parameter. Being biased, the sample estimate is not an accurate measure of the parameter. Even if an estimator is unbiased, errors of measurement and non response may produce biases in the numbers that we are able to compute from the sample data. This happens, for instance, if the persons who refuse to be interviewed are almost all opposed to some expenditvv?r, of public funds, whereas those who are interviewed are split evenly for and against.

Because of the difficulty of ensuring that no unsuspected bias enters into estimates, we will usually speak of the precidon of an estimate inctead of its accuracy. Accuracy refers to the size of deviations of the estimates from the true population parameter, whereas precision refers to the size of deviations of estimates from the limiting value of their mean as the sample size becomes infinitely large. /

Social science data are almost never precise or exact. Fmnomists and other social scientists have very imperfect measures and measuring rods oj the variables they study. In fact social science data are fundamentally les

than physical science data, since thcy are subject to additional inaccuracy in the measurement and reporting of human behaviour. Because of a variety of possible sources of error, including observation crror, round . off and approximation error, hiding of information, and errors of

national income measures involve quite high inaccuracies. In general two, three, or four figures of accuracy are all that can be expected of social science data

Population census figures are probably accurate to no more than the first two figures. If counting population is subject to large inaccuracies, measuring economic quantities and prices is subject to even greater ones. All items on any balance sheet or profit and loss statement of any individual or firm arc subject to inaccuracy. Many are based on arbitrary accounting conventions, others are measured subject to various hiases, and all are subject to reporting and other errors. Data on prices are particularly inaccurate, given various discounts, tie-in sales, quality considerations e t cetera which are typically not taken into account in price statistics. Social science data vary in their accuracy, but one has only indirect evidence or subjective opinion as to the precision of reported data since there is usually no indicatioil of their accuracy. This state of affairs contrasts from that of the physical sciences where in those cases in which experimentation is possible the differences in accuracy of measurement are indicated numerically by error bracket.

One more point to be made about economic data is that the errors in accuracy are generally not symmetric. There are often biases in one direction, so the error brackets are not equal in both directions. An example is total corporate profit as derived from corporate income tax returns. To the extent that corporations bias their profits downward to avoid paying taxes, the tot& figure may be subject to an error of say, 20% on the positive side but only 1% on the negative side.

The researcher should always keep the inaccuracies of the data in mind. Sometimes the unsatisfactory performance of a statistical model may be due to data iriaccuracies rather than the statistical teChnique. So rather than revisc the model or try a different technique, the researcher may investiqpt; &i

data more carefully and refine them further or else utilize altemsik data.

\ Scientific Research (Chap. 21 62 ,

variable customarily assumes a value of zero or unity, unity usually referring to the occurrence of an event or the presence of a characteristic, and zero referring to the non-occurrence of the event or the ribsence of the characteristic.

Dummy variables can also be used to distinguish among seueral qualitative characteristics or attributes, such as educational level, occupation, and region. If there are more than two possible characteristics or attributes, several dummy variables can be used. In general, a dummy variable is a numerical variable representing information of a qualitative type, typically showing the presence or absence of a qualitative characteristic. Use of such a numerical representation of qualitative factors facilitates the use of regression or econometric and other quantitative methods to analyze such factors. Indeed, if a regression uses only dummy variables as. explanatory variables, then it is equivalent to an analysis of variance, while if it includes both dummy and numerical variables as explanatory variables, then it is equivalent to an analysis of co-variance.

The use of dummy variables to express qualitative data is necessary only when the data are to be analyzed by regression or econometric methods. In this case, observations of a phenomenon are made over time (time series) or across space (cross section) and the value of the observation at each point in time or else at each spot or location is made to be equal to a value of the corresponding dummy random variable. Econometric methods are treated in Volume Three.

There is a multi-form device for distinguishing between observations of qualitative variables. Such a device is called scaling. Nominal scaling entails merely grouping the observations into labelled or distinct categories; that is, each category bears a name or label, which is indicative of the type of quality of each observation. Each observation bears one of a set of distinct qualities,

. attributes or characteristicssuch as ''yes," "no," and "not sowre," by which the - observations can be classified. If desired one could assign numbers to the

various groups or categories as a labelling device, but it would not matter which number was assigned to which category, and no meaningful calculations could be performed using these numbers. In other words nominal scaling is merely a way of classification rather than an arrangement along a continuum, and the question of dimensionality does not arise.

In some- other qualitative observations, there are gradations of each quality or attribute. All the observations possess the same attribute or characteristics, but the strength or intensity of the char~cteristic varies

between the observations. In this case, it is not intcndcd to categorise observations by type of quality, but to estimate the strength or intcnsity of the quality or attribute in each observation. The intensity of a quality is presumed to m r in a uniform gradation or scale stretching from the lowest extremity to the highest extremity. So, in this case, the observations can be ranked or manged in ascending or descending order of the intensity of their common attribute. In other words, the observations can be ranked based on a scheme such as highest to lowest. Such a ranking of observations is known as ordinal scaling, and it is used in attitude tests. For example a survey could be carried out to determine the attitude of consumers to a number of commodities, the objective being to rank the commodities in order of consumer's preferences. In this survey experiment, a consumer's scale of preference is ascertained to be in accordance with a scheme showing the degree of intensity of consumer demand as in this example: no demand, low demand, moderate demand, high demand, very high demand. On the other hand, the responients themselves (consumers) could be ranked according t o the intensity of their demand for a particular commodity, that is the intensity of the attribute possessed by the respondents. Ordind scaling merely ranks the subjects, but does not entail quantitative or numerical measurement and does not indicate magnitudes. In other words, the distance between any two rank positions is not a measure of the relative strengths of corresponding attribute intensities. We can only say that one rank position is greater than another, but not how much it is so.

Infewal scaling not only ranks observations in exactly the same manner as ordinal scaling, but also indicates how much one rank position differs from another. Measurements of observations must be such that they can be added or subtracted. So, interval scaling separates observations by rank order and measures the distance between rank positions in equal units; such a scaling method permits the researcher to say that position 6 on the scale is above position 7, and also that the distance from 7 to 6 is the same as the distance from 6 to 5. (Recall that the ordinal numbers are: lst, 2nd, 3rd, 4th, Sth, etc and the 2nd position, say, is greater than the 3rd position by ordinal scaling.\ However, interval scaling does not permit such conclusion as that position .i is twice as strong as position 8 because the zero position is not established! Such a comparison, however, is possible with mtio scaling. This is an interval scaling for which there is a zero position. In ratio scaling, measurements of observations can be added or subtracted, and multiplied or divided by real nuxbers (numbers that lie on a number line). These algebraic operations on the collected data will be meaningful only if the data themselves are

This is an exaqle of a discrete distribution of probabilities. In some cases, the numerical equivalements of the possible events form a continuum of numbers (the real line) and the distribution of probabilities are shown in characteristic graphs as in the case of normal distribution of probabilities - This is a case of continuous distribution of probabilities. In some researches, the adopted approach to statistical analysis of data does not require, any particular assumptions about the form of population distributions of probabilities. In this case the statistical methods employed are known as non- parametric or distribution-free,

2.6.4 Quantitative and qualitative data We have noted that sample data, whatever their bulk or reason for being,

are of two broad kinds, quantitative and qualitative. The former are measurements of some sort; they can be arranged in order of size. Qualitative data, on the other hand, do not allow an arrangement, but they can be categorized, grouped or classified by quality. A qualitative variable is also called an attribute. We have seen that quantitative data are either discrete (as a result of enumeration) or continuous (as a result of measurement). However, it is sometimes useful to treat both discrete and continuous data as qualitative and group them according to whether or not they fall into selected categories.

Sample data for which the distribution of pqobabilities is known can be analyzed or described by parametric methods. Qualitative data or categories of data and all other sample data for which the distribution of probabilities

. is either unknown or not required are described by nonparamebic methods. The latter do not require knowledge of distribution of probabilities and so the methods are distribution free. Both parametric and non-parametric methods use tables, charts, and graphs; and both attempt to reduce sample data to a few summay statistics such as measures of "central tendency" and of "dispersion." It should be noted, however, that such summary statistics as means, standard deviations, and other moment measures tend to lose meaning, even to be misleading, when the distribution of probabilities is not normal. .

Probability

?h

?h

. Events

Head

Tail

Numerical Equivalent

1

0

Examples of charts are the bar chart and the pie diagram. There are the contingency and frequency tables, and such graphs as histograms, polygons and ogives. These tables, charts and graphs are visual displays of the distribution of sample points by categories. For example, there are normal and skewed distributions. Knowledge of nature of the distribution of random variables is important in modelling situations. Modelling the processes or mode of generation of observations of situational phenomena enables statisticians to make inferences (that is, draw conclusions) about random events. Readers should be familiar with binomial, trinomial, multinomial, poisson, hypergeometric, gamma, normal, beta, chi-square, t, and F distributions. In more advanced research, the pareto distribution is frequently used as a model in the study of incomes, and the Weibull distribution is used in the study of breaking strengths of materials. Other less common distributions include the logistic, uniform, Dirichlet, etc.

There are certain common kinds of inferences or conclusions to be drawn about populations of observations. One inference has to do with the answer to the question whether respective effects of a given number of independent factors influence the outcome of a specific observation. For example, in the case of two factors, consider whether the variety of grain and the type of fertilizer used influence the yield; or whether the teacher and the size of class influence the score in a standard test. We must emphasize the necessity for independence of the factors. To answer the question, the researcher must consider the variation of the variable outcomes as the observations are subjected to the separate effects of the respective factors. The analysis of variance (ANOVA) technique is a statistical method used to determine if the separate factors have effects on the observations. Another question is whether a given pair of random variables are stochastically independent, that is, whether they do not covary, and such questions are answered after statistical analysis of co-variance between the variables.

An alternative approach to analysis of the relationship between a random variable and independent factors is regression analysis. Here a l l data must be in a quantitative or numerical form. Regression methods are treated in Volume Two.

Now, qualitative facts, for which no numerical measure exists, can also be expressed in the form of numerical data. This is done by expressing the qualitative data on appropriate dummy variables. The dummy variable takes one or two possible values, one value signifying one qualitative possibility and the- other value signifying the other possibility. By convention the dummy

- L. . rubllshed sources of data The federal government is a major collector of data for both public and private purposes. The Federal Office of Statistics (FOS), conducts annual labour force sample surveys on behalf of the National Manpower Board and the Federal Ministry of Employment, Labour and Productivity. The labour force survey is run as a module of the National Integrated Survey of Households (NISH) of the FOS. NISH is an on-going programme of integrated survey of households to collect data on household characteristics (social statistics) in such areas as income and expenditure, housing, health, and nutrition, employment, education, environment, agricultural production, etc. The programme consists of a set of core surveys that are run annually and supplemental survey modules on a less frequent basis. The FOS also conducts a survey of the distributive trades sector of the Nigerian economy, establishes the monthly Consumer Price Indexes, and publishes, in addition, demographic, labour, meteorological, vital, crime, educational, agricultural, foreign trade, travel, transport and communication, national accounts, and many other statistics covering various sectors of the economy. Much of FOS survey findings are published as special reports or can be located in such publications as the Annual Abstract of Statrrtics, Digest of Statistics and Social Statistics in Nigeria. The Central Bank of Nigeria (CBN) publishes statistics on the financial sector of the economy and these are contained in CBN Annual Report and Statement of Accounts, and CBN Statistical Bulletin. The Nigerian Stock Exchange publishers the Nigeria Stock Exchange Fact Book, which contains vital corporate statistics of quoted companies in the Financial, Manufacturing, Commercial and Services industries.

In addition to the federal government, various trade publications often present data pertaining to specific industrial groups, while individual annual reports often provide illuminating information on company-!eve1 activities. The daily newspapers often contain information regarding stock prices, weather conditions, and sports statistics.

. , The United Nations system publishes statistics on all member cmntries, which cover various sectors of the national economies, viz: national accounts, labour and financial, agricultural, and direction of trade statistics.

In all these cases, the survey experiment is performed by some external agency and the researcher merely uses sample statistics, which have a!ready been compiled, to conduct analysis of the given situation. To do so, he uses the method of statistical inference since the collected statistics are always assumed to be random.

2.6.2 Experimental data A second method of obtaining needed data is through experimentation (see 2.5.2.1 and 2.5.2.2). We already have explained business experiments and survey experiments. In our example in 2.5.2.2, a business experiment is performed by changing the business environment via an advertising blitz after which expected change in business activity is determined by sampling. It is possible to sample a population and then perform an experiment on the sample. Then by statistical methods, the behaviour of the sample points before the experiment is compared with behaviour after.

In all cases in which sample data are stochastic, inference about the population can be made @om sample information only if sample selection is randomized. Examples of sampling designs that ensure such randomization are given in 2.5.2.2. In the case of published data on the macro economy, business firms and institutions, the time series or cross-sectional data are assumed to be stochastic and satisfy the conditions necessary in order for use of statistical methods of data analysis to be applicable or valid.

2.63 Survey experiments We have already explained how sampling surveys are used to compare an "experimental situation" and a "control situatien." In many other cases such as "attitude tests" as in opinion polls, interviews, and questionnaires distribution, direct observations of phenomena or a phenomenon are made without any attempt to change the situation in which the observations are made.

To characterize the observations, the survey statistician wiil most likely want to develop an instrument that asks several questions and deals with a variety of phenomena. We have stated that the subject matter with which social scientists deal is random variables, which are phenomena of interest whose observed outcomes (data) may differ from response to response in a random or chance manner (i.e. unsystematically). Specifically, for each observation all the possible outcomes are known but the specific outcome that will be observed is not'lcnown. Now in statistical theory there are models of distribution of pmbability of obtaining outcomes over the entire range of outcomes. For example in the case of throwing a coin, the distribution of probabilities over the "events" (outcomes) are as follows:

Scientific Research {Chap. 21 56

generation of electric flow is created. This is achieved by devising an electric flow circuit. In such a situation, observations can be made on the effects of a sample of electrical cnerg in a particular circuit of electric flow.

In the physics experiment the cffect of electric flow is determined by contrasting the condition of experimental circuit with that of a "control" circuit through which current does not pass. Here the same effect is produced by all similar circuits. Measures of a particular effect may be read off directly from a reading device or calculated from data on observations. In the social sciences, however, a single observation is regarded as a random "event" or happening. For example, each one of a time series of sales, social or economic data or statistics, and students' scores in a test, is taken to be only one of the many possible alternatives. The one that is actually observed has occurred by chance, being only one of the probable observations or occurrences. A chance observation is only one of the possible observations of a random (chance or stochastic) variable. Recall that a variable is a phenomenon whose observations vary. For a random variable a particular observation is considered to be only one of the possible observations of which a specific one occurs by chance. Probability theory in statistics (study of statistics or data) is used to model the behaviour of probabilistic (chance or stochastic) phenomena. In sample surveys, care must be taken to ensure that sampling is conducted in a "random mannern so that a sample is actually a mndom sample. Now a sample measure or sample statistic merely describes the particular sample. A similar sample may yield a different measure, which is determined in like manner. The sample descriptive statistic cannot be used to make an exact general statement about the population. So, the physics experiment differs from an experiment in economics or management science because in the former a sample statistic (measure) is detem'nistic (non- random) while in the latter a sample statistic is stochastic (random). Accordingly only probabilistic statements can be made about the population when a random sample is used. Such is possible by means of the theory of inferential statistics.

There are a number of ways of randomizing.sampIes, that is, selecting them in a random manner. These include simple random sampling, systematic- sampling, stratified random sampling, and cluster sampling. (See Volume 2.)

252.2 Experimental research Certain researches (such as market testing) require the performance of an experiment. Once a problem is defmed with reference to a population, an

cxpcriment could bc pcrformcd on a segment of the population. An experimental research requires the delimitation of subscts of the population which are to be compared. Each of the subsets constitutes a distinct situation in which the phenomenon being observed exists. The conduct of the experiment is tantamount to the changing of the condition under which the phenomenon is observed in the "experimental situation." There must be a "control situation" in which the original conditions remain invariant. Given the changed conditions that are made prevalent in the experimental situation, a sample survey of the situation is carried out (i.e. survey research is conducted) in order to obtain relevant descriptive statistics which could be compared with those of the control situation. By method of inferential statistics it could be determined whether the experimkntal sample statistic and the control sample statistic are statistically different.

The method we have just described is used to evaluate the effectiveness of an advertisement blitz on urban consumers, say, the appeal of new product designs and marketability of new products.

1 253 Research into the tirture A researcher may wish to answer the question: What probably will be? Then he has to conduct a prognostic or predictive research. Such a research is conducted in order to extrapolate or prognosticate (forecast). Forecasting will be treated in Volume 4.

2.6 Research Data In section 2.3 we introduced the concept of data and referred to data collection and analysis as important ingredients in modem research endeavour. Indeed, the problem of data collection and analysis may be second in importance only to that of theory formulation and problem definition. Theory may originate from casual observations, repeated experiment or even intuition and a priori reasoning. However, sometimes theoretical postulates may arise from investigation or analysis of empirical or real world data. Recall that there are three basic methods by which the researcher may obtain - needed data. First, the researcher may seek data already published by governmental, industrial, or individual sources. Second, the researcher may design an experiment (see 2.5.2.2) to obtain the necessary data. Third, the researcher may conduct a survey (see 2.5.2.1).

1. UNITED NATIONS (UN) (1) UN

(a) Compendium of Social Statistics @) Demographic Year Book (c) Growth of World Industries me) (d) Human Report (e) Monthly Bulletin of Statistics (f) Statistical Yearbook (g) World Direction of Trade

(2) UNESCO Statistical Yearbook

(3) L.0 Yearbook of Labour Statistics

(4) WHO World Health Statistics

(5) F A 0 Production Yearbook

(6) IMF (a) Balance of Payments Yearbook (b) International Financial Statistics

itesearch Methods 299

APPENDIX 2

INSTRUMENTS OF DATA ANAYLSIS

CLASSICAL OR PARAMETRIC PROCEDURES 1. Research Objective: To compare two sets of observations (measures,

data). To achieve this research objective, it may be necessary to compare the respective measures of central tendency for the two sets, and/or the respective measures of dispersion.

The researcher may hypothesize that the two sets differ or else do not differ. If they differ, then their means must differ. To test the validity of the hypothesis, use Z or t statistic or any other suitable statistic.

Recall that if skewness of each of the respective distributions of the two sets is zero, you compare their arithmetic means. Otherwise, compare instead their medians. Test skewness of each of the distributions with Pearson's coefficient of skewness.

When the unit measures differ in scale, use the coefficient of variation to compare the variability of each of two sets of data. Other measures of variability (dispersion) include the standard deviation, the variance, and various measures of the range.

Two sets of data may differ because of some exogenous shock (disturbance) that may impact a research situation or because they are measures of observations of some phenomenon occurring between two different locations (places, zones, geogl~aphical areas, categories of some entity, etc) as in cross sectional analysis or else between two different periods of time as in time series analysis.

The mean and the median are preferred in measuring central tendency because confidence intervals can be established for them, and also because inferences about the population mean or median can be drawn from sample mean or sample median. The Z or t statistic is used tcj

establish confidence intervals for the mean, while the 0.5 order statistic, is used in the case of the median, which is the 50th percentile or quartile of order 0.5. (See also No. 9 and No. 12). When, however, the distribution of the sample points is skewed and the sample size is small, then if the researcher still wants to establish a confidence interval for the

- -4'PU- ' "-7 m~en he can use Chebyshev's inequality. The studenf m ~ u l d ~ i l i a r - i m s c ~ i t h tho ponditions that must -.

"-c-T1 hpfnrp. he can use the Z or t stat~st~c a-mu-Che,~v's

inequality. Further, he should familiarise himself with the' required conditions before two sets of data can be compared on the basis of

I

"difference scores" or "differences of scores." A research objective which entails comparison between two sets of

observations may involve test of differences of means or test of , differences of variance. The t statistic is the instrument used in testing

differences of means. To test a hypothesized value of the mean, use one-tailed Z or t test.

To establish a confidence interval for the mean, use two-tailed Z or t test or else Chebyshev's inequality.

When thc research objective is to determine which of two populations or two samples of observations has less variance (or has more uniform ,sizes) use the F statistic. When the research objective is to compare two populations by testing £or difference of variance, use the chi-square statistic.

Review the principle of the normal assumption in the test of difference of variance. You should be able to verify if a distribution is normal or

I not by, for example, testing for goodness of fit, using the chi-square I statistic.

The F statistic is used to compare the respective means of several I populations. Such a comparison is the basis of analysis of variance I (ANOVA) techniques.

2. Research Objective: To compare two sets of observathns (measures, data). Achievement of this objective may require comparison between two sample proportions, Such a comparison can be carried out in a number of ways, viz:

I

(i) If the sample size is small and the conditions of a Bernoulli process are met, use the binomial test for proportions.

(ii) If the requisite proportion is not too low (close to 0) or not too high (close to I) , and the population is large, use a large sample (sample size over 30), and provided the other conditions for a normal approximatian for the binomial are met, the Z statistic may be used. Under certain conditions, proportion problems are better solved by means of tdhe binomial or the Poisson statistic or else by the normal ay----.:-- - - for -r or"m.- -

Research Metlcodr 301

(iii) To estimate the proportion (percentage or probability) of some type of observation, use the binomial statistic to establish a confidence i n t e d for the proportion of the observations.

(iv) To test for differences among k sample proportions use the chi- . * square statistic.

Research Objective: To test for the compatibility of two sets of . distributions or. frequencies (e.g. observed frequencies and expected frequencies), or to test the goodness of fit of a hypothesized distribution of a given set 6f data with its actual distibution, use the chi-square statistic. An example is the research objective of comparing the distibution of attitudes to some subject held by two different samples of persons as in opinion polls or attitude tests.

Research Objective: To test the independence of two variables (two variable observations), use the chi-square statistic.

Research Objective: To investigate the effects of 2 factors that influence an outcome, use the F Statistic and the ANOVA technique involving two-way classification with one -observation per cell. The ANOVA technique and the F statistic are also used to investigate a two-way classification problem involving two or more stochastically independent observations per cell.

Research objective': To estimate the degree of correlation between two variables; to test a hypothesized value of the correlation coeficient between two variable observations, use the t, Z or W statistics.

Use of Chi-square tests: (a) To test the hypothesis that a sample comes from a particular

distribution (say the normal distribution). (b) To test the hypothesis that 2 distributions are the same. (c) To test the hypothesis that 2 distributions are'independent. (in all

cases sample categories must be exhaustive and exclusive).

Appendix 302

(a) Given a "Bernoulli process," determine the proportion of "successes" in a sample: use the binominal statistic;

(b) Given a "Poisson process," determine the number of "successes" in a fmed interval of time or fured interval of space, use the Poisson statistic;

(c) Given a "waiting time" situation, determine the waiting time till "death", the waiting time till "change" occurs or the waiting time till "success" obtains, use the gamma statistic.

9. Research objective: To compare several categories or groups by testing if their means are significantly different or by testing the null hypothesis that all the groups come from the same population by having the same mean, Le., the groups are identical by having the same means and hence belonging to the same population, ensure that certain requisite assumptions are satisfied and use the F statistic.

NON PARAMETRIC METHODS

10. In statistical inference it is usually assumed that the collected data constitute a random sample. Such an assumption may require a confirmation. So, to test whether collected data constitute a random sample or to determine whether an .observed sequence or sequences of collected data may be considered random, use the Wald-Wolfowitz one- sample-runs test for randomness.

11. To determine the existence of a trend in a set of data, use the Cox- Stuart unweighted sign test for trend.

To test a hypothesis regarding a parameter reflecting central tendency (espacially the median) in order to determine whether an observed value is larger or smaller than the hypothesized median or determine how much larger or smaller than the hypothesized median, use the Wilcoxon one-sample signed-ranks test, provided the necessary assumptions are met by the sample design. This is an alternative to the counterpart parametric approach, and is to be preferred under certain conditions (See Volume 2; also No. 1).

13. To test whether two mutually independent sample groups have been

Methods * ' 303

dram from the same population or identical populations, that is, to detect any kind of difference between the two groups - location, dispersion and/or shape, use the Wilcoxon ranks sum test. This is an alternative to the two-sample t test for differences of means. (See No. 1, where it is stated that the t o r Z statistic is used to test for differences of means). There are a large number of studies on consumer behaviour, ~lcuketing research, and experimental pyschology, in which it is required to compare two groups of ordinal type data. In such situations, the classical two-sample t test could not be used, and the nonparametric alternative must be employed.

14. In No. 13 above we considered test of differences between two independent groups. To test for differences among two related groups (as in test marketing a product), use the absolute normal scores test for related samples. This test is an alternative to the parametric t test for "difference scores" (a situation involving repeated measurements of the same item - See No. 1)) and "differences of scores" (a situation involving matched or paired items). The nonpararnetric alternative is preferred if the researcher does not wish to make the rigorous assumptions of the classical t test or else if the assumptions of the t test are violated.

15. In the classical approach, we considered the following researc.; objectives:

i) To determine whether there is any difference between the means of 2 independent populations. This required use of t or Z statistics to compare sample averages.

ii) To analyse the difference between the means of two related groups (not independent groups) as in the case in which the sampled items are paired or matched or else there is repeated measurements of the same set of sampled items or individuals - thc variable of intcrcs: being the difference between the values of the observations rathc~ than the observations themselves (case of "difference scores" or "differences of scores"). Use the t tests.

iii) If the researcher is not concerned with the differences between two populations in terms of a quantitative variable, but is interested in differences in some qualitative characteristic, then hc can test for differences between two proportions based upon indcpendcnt

I

In the case of qualitative data, the chi-square technique is used to examine differences in the proportions of "success" in two groups, and also to draw conclusions about the differences in the proportions of successes in more than two groups.

i n the case of quatitative data, there are methods that are developed to test for differences between the means of several groups. These are the "analysis of variance" (ANOVA) methods. These are the "one-waym ANOVA involving one "factor" (with several groups) or several factors with two or more groups. (See No. 5). All of them entail use of the F statistic. The F statistic is used to test for differences in the means of more than two distributions.

T APP& 304

-&s; he then can test, with a large sample, whether there is any difference in the proportion of "successesn in the N O p u p s (2- tailed test) or whether one group has a higher proportion of "successesn than the other group (one-tailed test), using the Z statistic. Alternatively, he can test for the difference between two proportions using the chi-square test.

iv) To test for independence of two distributions of "successesw. use the chi-square test.

v) To test for equality of two proportions of "successes", use the Z test or chi-square test.

vi) To test for differences of proportions of "successes" between more , than two independent populations, use the chi=square starktic.

~ i i j To te2t for independence between more than two populations, use the chi-square test.

1 viii To test for the goodness of fit of a set of data to a specific probability distri'bution, use the chi-square test.

ix) In analysing quantitative variables, such as the mean, it is often

I important to draw conclusions about the variability as well as the average of a characteristic of interest,

x) To test whether or not the population variance is equal to a specified value, use chi-square statistical test.

) To test whether two populations have the same variability, use the F test.

xii) So far, we have considered ((i) - (x) above) {he use of statistical inference to draw condusions about differences between two groups based on either a quantitative variable (mean) or a qualitative variable (proportions).

search Methods 305

5. Note: The Wilcoxon rank sum test is an malogut of the t test for two independent samples (See No. 15 (i)). In like manner, the Ktuskai-Wallis test is the analogue of the ANOVA F test (see No. 15 (xi)). To test for c independent samples (c > 2), thereby extending the Willcoxon rank sum test for two independent samples (See No. 13), use the Kruskal-V@lis test for more tban two independent samples.

university of nigeria enugu campus … tochukwu.t_08...faculty of business administration department...

Documents