Hi everyone -- here are some questions that arose in office hours (and our responses) that we thought would be helpful to share.
Question: How should we approach the idea of exchangeability in a case-control study?
Answer: The identifiability criteria (exchangeability, positivity, consistency) are independent of study design -- however, your study design can inform how you will approach each criterion. For example, the criterion of exchangeability relates to the broader concept of confounding bias (i.e., are your unexposed/exposed "exchangeable" or is your exposure-outcome relationship "unconfounded"?), which plagues all observational study designs. Strategies to address confounding bias, however, can differ depending on study design (i.e., in a case control study, matching controls to cases is one strategy to deal with confounding (assuming that your analysis adjusts for the matching factors)).
Question: For number 7, are you asking us to comment on potential differences in statistical efficiency/precision between the adjusted and unadjusted estimates?
Answer: Approach this question by comparing the adjusted/unadjusted measures of association. I.e., focus on the point estimates more so than the confidence intervals.
Question: Should we approach number 5 by comparing values in the tables?
Answer: Question 5 can be answered by reading the methods section. No need to examine tables for this question.
Happy homeworking! :)