On the robustness of weighted methods for fitting models to case-control data

We compare the robustness under model misspecification of two approaches to fitting logistic regression models with unmatched case-control data. One is the standard survey approach based on weighted versions of population estimating equations. The other is the likelihood-based approach that is standard in medical applications. The conventional view is that the (less efficient) survey-weighted approach leads to greater robustness. We conclude that this view is not always justified.