If your research question is about estimating an association, which association? If your question is to estimate the association between physical activity and diabetes incidence, for example, which association are you referring to? By using the word “the” you’re implying that there is only one or at least one that enjoys some kind of privileged position. But there are as many associations as there are variables to adjust for (and even more actually). So which association are you referring to?
You’ll find that, if your question is about estimating an association, it is impossible to say that one specific association (conditional on a specific set of variables) is the one you’re interested in. I argue that the only way to say that one specific association is the one you’re interested in, you have to appeal to a better research question. If your real research question is causal then the association you’re interested in is the one you would observed if there were no confounding, selection bias, measurement error etc. If your interested in a descriptive question, that question will define your adjustment set. For example, you might want to know about the inequality in mortality between migrants and non-migrants but without considering the difference in age structure. Then you would adjust for age (in some way).
You might try to wriggle out by saying that you’re interested in the association given the covariates that you chose to adjust for in your analysis. But how did you choose those covariates? There’s no way to decide what to adjust for unless you have some clear question.
So just don’t use the word association in your research question. I sound like I’m grumpy here but I’m mostly not.
