chapter 12 4/11/2013 11:261geog 3250 baxter. sampling - exercise what is the goal of sampling? how...

Chapter 12 4/11/2013 11:261Geog 3250 Baxter

Sampling - Exercise What is the goal of sampling? How do sampling goals differ between qualitative research and quantitative research? 4/11/2013 11:26Geog 3250 Baxter2

Why is sampling important? In 2010 the mandatory long form census was abolished and in 2011 was replaced with the voluntary National Household The controversy was over sampling! 4/11/2013 11:26Geog 3250 Baxter3 Click image to watch

Key Terms Element or unit: a single case in the population Population: all cases that a researcher is interested in Sampling frame: the list of elements that the sample will be selected from Sample: the elements (subset of a population) selected for investigation 4/11/2013 11:264Geog 3250 Baxter

Key Terms, contd Representative sample: a sample that contains the same essential characteristics as the population Probability sample: a sample selected using a random process so that each element in the population has a known likelihood of being selected Non-probability sample: a sample selected using a non-random method 4/11/2013 11:265Geog 3250 Baxter

Key Terms, contd Sampling error: estimation of the error that occurs because of differences between the characteristics of the sample and those of the population (more on this below) Non-response: when an element selected for the sample does not supply the required data Census: data that comes from an attempt to collect information from all elements in the population 4/11/2013 11:266Geog 3250 Baxter In 2011 the short form census remained mandatory, while the long form census did not (this became the new National Household Survey

Sources of Bias in Sampling Not using a random method to pick the sample The sampling frame Human judgement that selects one group over another Non-response Some people in the sample fail to participate which skews the data. 4/11/2013 11:267Geog 3250 Baxter

Sampling Error Errors of estimation that occur because there is a discrepancy between the sample group and the total population. Virtually impossible to eliminate sampling error. Using random samples and making the sample large helps to minimize sampling error. 4/11/2013 11:268Geog 3250 Baxter

4/11/2013 11:269Geog 3250 Baxter

Types of Sampling 4/11/2013 11:2610Geog 3250 Baxter Click image to watch

Types of Probability Sample Simple random sample Each element has the same probability of being selected. Each combination of elements has the same probability of being selected. Known sampling error for each sample size allows us to make probabilistic conclusions (more about this later) 4/11/2013 11:2611Geog 3250 Baxter

Types of Probability Sample, contd. To select a simple random sample: Devise a sampling frame A list of elements in the population Number all the elements consecutively stating at 1 Pick a sample size (n) from the total population (N) Using a random number table or computer program to generate a list of random numbers The sample will be comprised of the cases whose element numbers match the randomly generated numbers 4/11/2013 11:2612Geog 3250 Baxter Create random list of integerslist of integers Randomize a text list (cut and paste)text list

Types of Probability Sample, contd. Systematic sample: every ith case in the sampling frame is selected. Simpler to execute (ironically) than simple random sample i = size of sampling interval e.g., if you want to select every 3rd case, i = 3 To begin, choose a number at random from 1 to i, e.g., 1 to 3 = random start Then select every ith case after that (in our case, every 3rd case after that). 4/11/2013 11:2613Geog 3250 Baxter Random number from 1-3, picks 1 4 7 10 13

Types of Probability Sample, contd. A potential problem with systematic sampling: periodicity This occurs if the cases in the sampling frame are arranged in some systematic order, e.g., in an election study: voter, non-voter, voter, non-voter, etc. If we were to select every 30th case starting with case 20, we would select case 20, 50, 80, etc., i.e., all cases would be non- voters E.g., odd and even numbers in street addresses 4/11/2013 11:2614Geog 3250 Baxter odd #seven #s Sampling interval of 2 = only one side of the street and side-by-side neighbours may think and act in systematically different ways than across the street neighbours

Social Science Types of Probability Sample, contd. Stratified random sampling Ensures subgroups in the population are proportionally represented in the sample. e.g., assume you want a sample of 100 students and want to ensure that each faculty is represented in the sample proportionally. May be hard to get disaggregated lists Better than simple random sample, will likely reduce sampling error for parameter that is basis of stratification 4/11/2013 11:2615Geog 3250 Baxter 400020001000 Science Business 7000 Sample Frame 572914 Stratified random sample of 100 from 7000

4/11/2013 11:2616Geog 3250 Baxter Simple random vs stratified random sample of 450 students we assume stratified is gold standard e.g., 5 too few in simple random e.g., 20 too many in simple random Advantage of stratified vs simple random sample

Types of Probability Sample, contd. Multi-stage cluster sampling Used for large populations. No adequate sampling frame Elements are geographically dispersed. It involves two or more stages. Selecting clusters (groups of elements) Then selecting subunits within clusters 4/11/2013 11:2617Geog 3250 Baxter A group of Western students may be considered to represent all students in Canada using mscs

Example Random cluster sample of 1000 Canadian adults Randomly select clusters, e.g., five provinces or territories from a list of all provinces and territories. Randomly select e.g., four census tracts from each province/territory Randomly select 50 households from each census tract selected (i.e., 1000 divided by five, then divided by four). Randomly select one person to be included in the study from each selected household (e.g., closest adult birthday to survey date). 4/11/2013 11:2618Geog 3250 Baxter

Example, contd. Technical complications, e.g., not all clusters are the same size Antidote cluster at the highest level in the chain In our example, the provinces and territories might be categorized into regions: BC, Prairies, Ontario, Quebec, Atlantic provinces, and the northern territories. Select the first clusters proportional to these regions Then the next steps would be taken. 4/11/2013 11:2619Geog 3250 Baxter

Sampling Error 4/11/2013 11:2620Geog 3250 Baxter This video nicely summarizes the difference between variation and error and the sources of each

Sampling Error Probability samples with sufficient sample sizes minimize the amount of sampling error, but some sampling error will occur. e.g., there is usually some difference between a sample mean and the population mean () that it is designed to represent. This sort of sampling error is measured by a statistic called the standard error of the mean (which you will never be asked to calculate in this class) 4/11/2013 11:2621Geog 3250 Baxter

Sample Size The absolute size of the sample matters (not the proportion of the population that it comprises) As sample size increases, sampling error tends to decrease. Common sample sizes:100, 400, 900, 1600, 2500 Each size increase cuts the sampling error by 1/2, then 1/3, then 1/4, and then 1/5 respectively. The biggest change occurs between 100 and 400. Is an increased sample size worth the time and effort? Often sample size is dictated by financial concerns. 4/11/2013 11:2622Geog 3250 Baxter

Non-response The response rate is the percentage of the sample that participates in the study. If there is some particular issue common to the non- responders that brings them to differ in some important way from those who participate. 4/11/2013 11:2623Geog 3250 Baxter Table 3 2011 National Household Survey (NHS) and 2006 Census long-form response rates Provinces and territories 2006 Census long- form final response rate (%) 2011 NHS preliminary response rate (%) Canada93.569.3 Newfoundland and Labrador 94.364.0 Prince Edward Island 93.761.5 Nova Scotia93.365.8 New Brunswick95.164.5 Quebec93.672.9 Ontario94.368.1 Manitoba95.568.8 Saskatchewan95.164.5 Alberta93.767.5 British Columbia91.670.2 Yukon96.061.1 Northwest Territories 97.683.9 Nunavut93.878.1

Heterogeneity of the Population Generally, the greater the heterogeneity of the population on the characteristics of interest, the larger the sample size should be. 4/11/2013 11:2624Geog 3250 Baxter

Kind of Analysis The sample size needed my vary depending on what sort of analysis will be done (see analysis and interpretation lecture I: quantitative analysis) If small groups in the population are to be compared to larger groups, it may be necessary to oversample the smaller group in order to make meaningful comparisons. Certain statistical procedures, such as some multivariate analyses, require large sample sizes to work properly. 4/11/2013 11:2625Geog 3250 Baxter

Types of Non-probability Sampling Convenience sampling Cases are included because they are readily available. e.g., one could go to a mall and administer a survey to anyone willing to take part. Problem: one cannot generalize the results to some larger population with any confidence Convenience samples are useful for pilot studies, for testing the reliability of measures to be used in a larger study, for developing ideas, learning how do to research as in this class 4/11/2013 11:2626Geog 3250 Baxter it will only take 2 minutes

Types of Non-probability Sampling, contd Snowball sampling a unique form of convenience sampling Very useful for finding people for which there is no conceivable sampling frame The researcher makes contact with some individuals, who in turn provide contacts for other participants e.g., member of activist group puts you in contact with other activist group members 4/11/2013 11:2627Geog 3250 Baxter This is a metaphor: It is unethical to roll participants inside snowballs

Types of Non-probability Sampling, contd Quota sampling Collecting a specified number of cases in particular categories to match the proportion of cases in that category in the population. e.g., there quotas for people in certain groups such as age, gender, ethnicity, class, etc. Random methods are generally NOT used to fill the quotas (as it would in stratified sampling) 4/11/2013 11:2628Geog 3250 Baxter

Quota Sampling Quota sampling - is used a lot in market research, but is rarely used in social scientific research in North America Criticisms: Quota samples are not likely to be representative. Judgements about eligibility may be incorrect,. e.g., a researcher may misjudge a persons age and mistakenly avoid the person The data gathered cannot be used to calculate inferential statistics (based in random sampling). 4/11/2013 11:2629Geog 3250 Baxter

Quota Sampling, contd Strengths of quota sampling: Cheaper, and easier to manage compared to random sampling Can be conducted much more quickly than random sampling Good for pilot tests, exploratory research 4/11/2013 11:2630Geog 3250 Baxter

Sampling in Structured Observation Often no sampling frame e.g., a list of all people who were admitted to the emergency room at a particular hospital May involve time sampling e.g. an emergency room may be observed at random times throughout the day May include place sampling e.g. a study of student activities on campus may involve a sampling of places such as dining halls, pubs, classrooms, etc. May include be behaviour sampling e.g., a researcher may want to observe every fifth interaction between students and librarians at a particular reference desk 4/11/2013 11:2631Geog 3250 Baxter

Content Analysis Sampling Media may be sampled: e.g., a study of newspaper articles may involve sampling of different papers, of articles on a given topic, etc. Dates may be sampled: e.g., if researching media portrayals of prostitutes, one could use a random method to select the years for which the media are to be analyzed 4/11/2013 11:2632Geog 3250 Baxter

Qualitative Sampling In ethnography, convenience sampling and snowball sampling are commonly used. Some qualitative researchers engage in theoretical sampling this research is inductive after all! Data are simultaneously collected and analyzed. Data collection is determined by whatever theoretical or conceptual issues emerge as the study progresses. 4/11/2013 11:2633Geog 3250 Baxter

Qualitative Sampling, contd. In addition to people, times and contexts may be sampled in qualitative studies. e.g., if observing a biker gang, different times of the day should be used, as well as different contexts, such as those involving the presence of rival gang members, the presence of law enforcement officers, etc. 4/11/2013 11:2634Geog 3250 Baxter

Sampling Problems Sampling error (addressed above see sample size) Sampling related error Arises from activities or events related to the sampling process, e.g., non-response, inadequate sampling frame, etc. 4/11/2013 11:2635Geog 3250 Baxter

Reducing Non-response Call backs are useful. Sometimes several are necessary. They reassure prospective participants that you are not out for material gain. Dress appropriately for face-to-face contact. Be flexible to accommodate participants. 4/11/2013 11:2636Geog 3250 Baxter

chapter 12 4/11/2013 11:261geog 3250 baxter. sampling - exercise what is the goal of sampling? how...

Documents