Usually, you want to control variables within a study. However, you also want to obtain a representative population to the population that you're measuring. How can one even strike a balance between the two? Is it by splitting the population into specific groups with the specific RELEVANT control variables, as you specified in this link (Is controlling variables important in deducing a statistically significant correlation?), and then observe the relevant outcomes?

I think that representation is also confusing because because I think to myself just because certain people live within a neighborhood, is there necessarily a link between them? There are probably certain aspects of representation which we have not been able to study yet.

  • 1
    $\begingroup$ This read might address some of your concerns: Why representativeness should be avoided $\endgroup$ May 8, 2017 at 18:53
  • $\begingroup$ In the related question, you've asked me to look at this one as well (at Mathematician), I have, and the link @ManuelFazio provided provides an excellent and also detailed answer to how representative your study population has to be (or not). A short example I can readily give is the following: the 'god' amongst studies types, the RCT almost never has a perfectly representative study population. As in, due to the strict in- and exclusion criteria, specific patients groups (which might be the most important), often get left out (think of children, pregnant women, elderly, etc.). $\endgroup$
    – IWS
    May 8, 2017 at 19:46


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.