sampling issues

35
Slide 1 Slide 2 Slide 3

Upload: alan-p-jack

Post on 19-Nov-2014

932 views

Category:

Business


2 download

DESCRIPTION

 

TRANSCRIPT

Page 1: Sampling Issues

Slide 1

Slide 2

Slide 3

Page 2: Sampling Issues

Slide 4

Slide 5

Slide 6

Page 3: Sampling Issues

Slide 7

Slide 8

Slide 9

Page 4: Sampling Issues

Slide 10

Slide 11

Slide 12

• Sampling Issues in Quantitative Research

Anji Waring

Page 5: Sampling Issues

Faculty of Health and Social Work

What is a sample?• A small group drawn from

a larger population.• You could to it on an

entire population but it is unusual

• Looking for a group who represents the group

• The population is the entire set of subjects in a given group that form the focus of the study.

• The group you are going to apply the results

Page 6: Sampling Issues

• Want to apply the results to the bigger group

• May want to limit it (uk population of women over 50 with breast cancer

Who What When Where

• It may be necessary to distinguish between the theoretical and accessible population

IN THEORY WHO CAN YOU ACCESS (THEORETICAL) THOSE CAN GET ACCESS TO YOU (ACCESSIBLE) NOT BE ABLE TO GET TO EVERYONE!!!

Page 7: Sampling Issues

How can you access your sample?

• Sample Frame: A list, register, map or other set of data that contains all the accessible population. (e.g phone book; electoral roll; NMC register)• HOW YOU ARE GOING

TO CHOOSE YOUR SAMPLE IN THE FIRST PLACE!!!Not always obvious from the surveyNeed to read and find out how they got hold of the people.

Page 8: Sampling Issues

• The sample is the group selected from your sample frame – not the group who are actually in the study.

Some may say no – they are still part of the sample because they become the non-

respondants. So sample is the numbers who did and didn’t respond Beware of self selecting they can produce bias. Poor response rate could mean you are missing out on poeple

Selecting a sample• In order to draw

conclusions about a larger population, the sample must be representative of that population.

Got to make sure that it respesents the numbers and the types of people studied.

• There are two main approaches to sampling: Probability (proably

Page 9: Sampling Issues

represattive) and Non-probability (probably not resprestative)Probability Sampling

• Any method of sampling that utilizes some form of random selection. This means that different units in your population have equal chance of being chosen.

Like raffle The lotery

• This is often done by using random numbers either generated by computers or by using tables.

Page 10: Sampling Issues

• Usually more structured in research

See hand out.

One hundred people who attended an outpatient clinic.

So you have 00 to 99 just the way of making sure that there is an equal cance of being selected. (Cormack (2001))

You can use random numbers and do things lk usisnt them backwards or

Page 11: Sampling Issues

upside down or inside down. Would this give you a random and representativeA way of not being biasedSome use computers – work it out.Sample to respesent the population as a whole.

Look for demographic tableBreaking it all down Then you decide if that is

represtative of what you are looking for.

Page 12: Sampling Issues

Types of probability sample

• Simple Random sampling: Generated from random tables etc of the whole population.

• Take the whole population. Take the sample frame reprenstitive of population then sample randomly from it.

• Accepted as the best way!!!!• Stratified Random sampling:

(proportional or quota random sampling) – dividing the sample frame into homogenous subgroups and taking a simple random sample from each group.

• Cluster Sampling – divide population into clusters (e.g

Page 13: Sampling Issues

wards) and then take a random sample of the clustersMulti Stage Sampling: Cluster then stratify then randomise

Although you do a random sample, you need to know the demography because chance can result in a bias in the first place.Looking to build a body of evidence. Evidence that would help to inform rather than just as a one of piece of research (eg, MMR)

Non-Probability Sampling

Page 14: Sampling Issues

• Includes all sampling procedures in which chance plays no rule in the determination of the actual make up of the sample.

Less rigourousVery often doneThe easiest way to getting a sample

Some include all the sample frame even though it is not a randomly generated sample.

Page 15: Sampling Issues

More descriptive but still quantitive.

Types of Non-Probability Sampling

• Convenience Sampling: Based on accessibility to the researcher rather than on the basis of random sample procedures. Often used when time and resources are limited.

The weakest form of samplingBut it is the most commonly used

Page 16: Sampling Issues

Not generalisable to a very great extent.Sometimes, there is some randomisation used

Make for more rigour but still not good enough• Volunteer Sampling:

Sample consists of subjects who have responded to an advertisement & have volunteered to take part in the study.

You are not trying to generalise

Ie, survey want people with a bereavement difficult to get a list of

Page 17: Sampling Issues

people ask for volunteers got to a place where you know you will find people.

You only get people who have axes to grind.

Journal samples violence in Nursing standard in 1986 just printed in the journals

400 odd people responded 0.05%

78% of the 0.05%Then they said that 78% of

nurses have suffered violence!!!!

Page 18: Sampling Issues

Mostly men (men are more likely to be hit as they work in places where it tends to get hit anyway)

But it was a volunteer sampleThey may have an axe

to grind It may be the only way

to access that groupWhat type of sample If it was a volunteer

sample, was effort made to make it representative.

Page 19: Sampling Issues

Types of Non-Probability Sampling

• Quota Sampling: Deliberate choice of approaching a quota of respondents to represent the population (e.g. men and women)

Done in market researchThey have to have x who are

y, x who are z… etc.

There is some chanceBut not an EVEN chance!!!!

Who ever happens to be thereNot a true sampleA deliberate choice

Page 20: Sampling Issues

• Snowball: In hard to reach groups, original respondents are asked to name others who share their characteristics

Ie IDVUS: who are part of a needle exchange. “could you tell your mates…..” Yes…. But….

Not randomMay not be trueConfidentialityCould be people who are

similar with reduced diversity(did it with football hooligans)Everyone in the same gang had similar views

Samples tend to be smallThere are skews There are biases

Page 21: Sampling Issues

• Purposive Sampling: Term used in qualitative research.

Gone out looking for people who have the attributes that you are looking for. Not randomJust for QUALATIVE!!!!!

Sample SizeMore important in

quantatitve rather than in qualitive….

Going on and on until you reach saturation

Where you aren’t getting anything new• Has to be large enough to

allow for generalisation.

Page 22: Sampling Issues

This is influenced by:– Subgroups for data analysis

need to be representative of the smallest groupIs this big enough for this to be valid….A statistision can usually tell you.

– Effect size – difference Should be worked out for you by the researchSignificant numbers in the group for it is doing what it is purporting it is.

Page 23: Sampling Issues

– between groups– Statistical calculations

(may require a power calculation to determine size)

Gives you a number of how many you need.Need to look at if they say that a power calculation of how many people you need so that they need to get to make it do what they want it to …The bigger the better!!!!!!

– Likely response rate

Page 24: Sampling Issues

Postal research rates tend to be lower.If you want 100, you need to send out 4-500 Face to face, it is less but you still need more as some will not be interested and also you will have biases due to the interviewer used…..You may have information about theose who didn’t respond.

Page 25: Sampling Issues

Keep a demography of the non-response rate.Careful how you use the information. is this reported without there consent.No mimium size but the smaller the size, the less represntable thus the less generisable.

Bias in relation to Sampling

Page 26: Sampling Issues

• Sampling Bias: means that the sampling procedure results in a sample that does not represent the population of interest.

• Selection Bias: occurs if the characteristics of the sample differ from those of the wider population.Randomisation of a Sample (Random

Assignment)• This is NOT the same as

random sampling.Random sampling is when you randomise

Page 27: Sampling Issues

Randomisation is when you have your sample, you then randomly assign them to a group…

So you can have a convenience sample and then randomly assign them.

• A procedure which is used to assign subjects randomly to treatment or control groups, in which the subjects have an

Page 28: Sampling Issues

equal opportunity to be assigned to either group.