6
$\begingroup$

As I understand it, one of the main goal of the chi-squared test on a contingency table is to determine if the link between lines and columns of the table is "more" than the sampling bias and the random fluctuations it can generate.

So my question is : if a contingency table contains the whole surveyed population, there is no more sampling bias. So does it make sense to apply a chi-squared test on this table, or do we just have to look at line and column percentages without worrying to test against the null hypothesis ?

Thanks in advance for any hints !

$\endgroup$
7
$\begingroup$

If the whole population of interest has been surveyed, then there's really no statistical analysis to do - any differences you see are differences in the population by definition.

$\endgroup$
  • 2
    $\begingroup$ I would amend @EpiGrad 's answer to say there are no inferential statistics to be done. You can certainly do descriptive statistics. $\endgroup$ – Peter Flom - Reinstate Monica Sep 20 '12 at 9:52
  • $\begingroup$ There is a theory in survey sampling of superpopulation models. If you assume that the population is a sample from a superpopulation then I think there is a way to do statistical inference. But to apply such a model would mean that you are interested in answering a different question than whether or not the proportions are different in this given population. $\endgroup$ – Michael Chernick Sep 20 '12 at 11:14
  • 2
    $\begingroup$ Here is a link to a nice introductory article on superpopulation models this $\endgroup$ – Michael Chernick Sep 20 '12 at 11:18

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.