What is target and sample population? the portion of population for whom we have received data or those who were invited to participate

Question

I am new to statistics and I am trying to understand target population and sample data.

Suppose a company wants to know the level of satisfaction of all their customers. To this end, it has sent a survey questionnaire to a portion of its customers (not all) to know about their satisfaction. My question is what is the sample data and the target population in this case.

For sample data, would it be all the customers who have received the questionnaire? Or would it be those who received the questionnaire and completed the survey? I am inclined to say those who have provided a response, because I think sample data is concerned with the available information, so if someone received a questionnaire but chose not to participate would not be part of the sample data.

For target population, I think it would be all their customers, including those who haven't received an invitation to participate in the survey. But I can also see someone can argue the target population would be those who have received an invitation, and the sample data would be those who have responded.

In general, is there a definite answer when statistician talk about target population and sample population or is it more of a subjective matter?

Answer 1

The target population is the population that is your goal for whatever it is you are trying to estimate. Ideally you would measure everyone in the population and that would give you an exact value for customer satisfaction. But as you usually cannot measure everyone, you settle for just taking a sample.

The sample is only the set of data that you collected (those that were both sent the survey and responded).

Suppose your measure of customer statistication is a simple binary yes/no. The proportion of 'yes' in the target population is called a parameter. It is what you are trying to estimate from your sample.

So the issue is was the sample obtained in such a way that the proportion of 'yes' in the sample is likely to be a good estimate of the parameter (the proportion of 'yes' in the target population)? Here the answer is probably no because those who choose to respond may be either more or less likely to be a 'yes'.

An alternative way that one might define things is to refer to the group of people who were sent the survey as the sample, but... importantly, it has missing data in that there are a group of people who did not respond. This then gets into difficult issues that, as someone new to stats, you should not try to make sense of now.

So if you are going to stick for now to the simplest way to define these concepts, the sample is always only the data you actually have.