Let's assume I want to conduct a survey. To find test subjects, I post an ad in a local newspaper, one on a locally relevant website, distribute flyers in front of the train station, and have managed to get one complete class from a local school to partake.

Due to the different recruitment methods, each will probably have drawn quite different samples from the population: the local newspaper is mostly read by older people with medium to high education; the institute website is mostly visited by students and researchers; the flyer is given to commuters; then there is that group of pupils, all of the same age and education level.

In my test, I ask my test subjects to specify how they knew of the test:

  1. ad in newspaper
  2. ad on website
  3. flyer
  4. they were tested in class

So I can group my data according to the recruitment method, if I want to.

Is this all one sample? Or have I drawn four samples?

And if this is one sample, how then do you draw two or more samples?

  • $\begingroup$ "Let's assume I want to conduct a survey." Are you really? It looks a lot like you're doing homework. $\endgroup$
    – Ana
    Jun 11 '13 at 11:43
  • $\begingroup$ It's not homework, but does it matter? $\endgroup$
    – user3116
    Jun 11 '13 at 12:46
  • 1
    $\begingroup$ I think it's an interesting question. I agree that it reads like homework. @what given that it's not homework, then I congratulate you on writing creative questions to learn material. As a side point this site hasn't received a lot of homework questions, so we haven't really developed a policy on the topic. Personally, I think the general problem with homework questions is that they tend to be too specific (i.e., they don't create a general resource for the internet). In contrast, I think this is a general question. $\endgroup$
    – Jeromy Anglim
    Jun 11 '13 at 13:23
  • $\begingroup$ @what - it might matter. My husband had a problem in a course of his when some homework questions got posted on one of the stackexchange sites, and many students copied the answers. The homework was supposed to carry points and be part of the final grade and all that. It would be good to know that answering your questions won't lead to similar problems for someone. $\endgroup$
    – Ana
    Jun 11 '13 at 13:33
  • 2
    $\begingroup$ This depends on what you are looking for. If you used the different recruitment methods to specifically draw from different populations, you can also consider them distinct samples. In general, it mainly depends on your definition what a sample constitutes in a particular case. You can recruit random people as they walk through the door and decide the first 100 to be your Sample 1 for Study X and the second 100 to be your Sample 2 for Study Y, but you could also call all 200 of them to be just one sample if it suits your needs. $\endgroup$
    – H.Muster
    Jun 12 '13 at 10:27

This depends on your research question, as well as the population you want to generalize your conclusions to.

If you care about whether complex sounds are easier to discriminate when they are longer, you can regard all the participants to be one sample because there is no theoretical reason to assume that people differ in this. In principle, the more universal the subject of your study is, the less you need to care about your sampling technique because any person should be as good as any other. In practice, the sample will likely be first year psychology students :)

But if you care about computer literacy or car ownership or attitudes towards standardized testing (i.e. if your question relates to individual differences or group differences), then clearly your samples are drawn from very different parts of the population. In that case you can say that you have drawn four samples. Or perhaps less - perhaps you drew two: people unlikely to have cars (school kids and train takers) and people likely to have cars (American newspaper readers and internet users). This is not the best example but you can see the logic.

  • $\begingroup$ Thank you, @Ana, this totally makes sense. I was so focussed on the recruitment, that I forgot to consider the purpose of the study and that my planing of that recruitment should relate to this purpose. This is a valuable insight: that everything, from the choice of test administrator to the daytime of the testing, must be put in relation to what I want to research. $\endgroup$
    – user3116
    Jun 12 '13 at 11:59
  • $\begingroup$ @what - Yeah, looking at research from the standpoint of someone planning to do it is different from looking at it in vacuum, i.e. methods courses :) $\endgroup$
    – Ana
    Jun 12 '13 at 12:35

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy