I'm having a difficult time understanding the difference between target population and sampled population using stratified sampling. For example, say I want to run some statistics on university students in the UK -- in particular, I want to know specifically about the first-year students. However, let's say the male/female ratio is 70/30, so I want to stratify the samples and take 1000 students from each gender group.

If the target population is defined as the "whole group of interest," does that mean all university students in the UK? If so, then I'm inclined to say the sampled population are university first-years, and the sampling frame is the 2000 students (1000 male/1000 female) chosen from my strata. However, I keep going back and thinking the target population is actually the university first-years in the UK, not the entire university population (first, second, third, fourth years).

Any guidance or tips would be appreciated.

  • 1
    $\begingroup$ As I understand it the target population and the sampled population are the same. You define the population of interest and then sample from it by taking a subset. For statistical inference the sample will be a form of random sampling. $\endgroup$ – Michael Chernick Mar 10 '18 at 1:16
  • 1
    $\begingroup$ For a statistical purpose, the population could be termed as finite or infinite. $\endgroup$ – Subhash C. Davar Mar 11 '18 at 5:06
  • $\begingroup$ @MichaelChernick I've added an answer, but I think we disagree on whether target and sampled population differ $\endgroup$ – RoryT Mar 16 '18 at 0:36
  • $\begingroup$ Yes a population can be finite or infinite. $\endgroup$ – Michael Chernick Mar 16 '18 at 1:15

The target population of a survey is the population you wish to study. The sampled population is the population which you are able to observe in a sample. In an ideal world the target population and the sampled population would be the same, but often they are different.

For example, governments around the world survey active businesses to determine estimates for labour force, productivity, innovation, capital expenditure, etc. Typically this is done by maintaining a register of businesses and sampling from this register. However, there is a gap of weeks/months between when this sample was created to when the survey goes to field. This means there are businesses who have folded after being selected for the survey (ie these entities are no longer in the target population because they have become inactive) and there are new businesses who have started and cannot be surveyed (ie these entities are not in the survey population even though they are active and in the target population). This difference needs to be accounted for, or else we would never be able to include new businesses in these economic measures.

In your case I would keep in mind:

  • Your target population is theoretical, your sample population is practical, based on the frame you have access to or are able to construct.

  • A target population should be more specific than what you've described. Specify exactly the units you mean: for example, full time students studying courses at universities located in the UK on 1st September 2018, excluding those studying by distance not residing in the UK. Note that the specification should be CUTE: content (who the survey is about), units (who you are approaching), time (when the survey is about), extent (where the survey is about)

  • The survey and target populations refer to the populations for the entire survey, not individual questions. If you want to ask general questions to all university students, but have some more specific questions for first year students then your survey and target populations are both still all university students, but you may want to control how many first years you survey through stratification to achieve the required accuracy for those questions that specifically concern them. If however you are only interested in first year students but you're sampling other students too, then your target population is just a subset of your sample population and you should try to fix this.

  • $\begingroup$ target population can be understood as a sub-population and it can not be considered as one " ... you want to study" . Usuallly; for a survey, it is impossible to specify and study the whole and exact population. So the population is targeted . $\endgroup$ – Subhash C. Davar Mar 11 '18 at 5:17
  • $\begingroup$ I have to disagree @subhashc.davar - a target population in the context of survey sampling is defined as the theoretical population you want to study. Here's a quick article: theanalysisfactor.com/target-population-sampling-frame You seem to be confusing the sample with the target population. The sample is the group of units actually selected to represent the target population $\endgroup$ – RoryT Mar 11 '18 at 6:42
  • $\begingroup$ @subhashc.davar - when you vote an answer down you lose one of your own reputation $\endgroup$ – RoryT Mar 11 '18 at 6:43
  • $\begingroup$ A downvote costs 2 reputation points. I think we are too rapped up in semantics on this problem. I suppose you can have a population that you want to draw inference from but can't sample in a way that all cases in that population are possible to be sampled. In that case you can't take a sample and properly draw inference about that "target " population. $\endgroup$ – Michael Chernick Mar 16 '18 at 1:22

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.