I'm conducting a case-control study consisted of 32 males and 35 females matched by age and gender with controls. Do I have to control for gender when I do the statistical analysis?


1 Answer 1


If you are analyzing each matched pair, then no, you should not include the matching variables in the regression model, but you should be using conditional logistic regression.

If, instead, you are using matching just to ensure a "balance" of covariates between cases and controls, you can help address some additional residual confounding by including those matching variables in the model. You do not have to do so, but it protects against failures in the matching process.

You may find much of this presentation helpful: http://www.teachepi.org/documents/courses/fundamentals/Pai_Lecture8_Confounding_Part2.pdf

  • $\begingroup$ thank you very much your answer is very helpful! what i understood is that when i'm conducting linear regression analysis for the full cohort i have to control for gender. And i have to use conditional logistic regression to obtain OR?-@Fomite $\endgroup$ Jan 5, 2016 at 16:25
  • $\begingroup$ @ReemM.AlHaj You should be doing one or the other, not both. I'd use conditional logistic regression and keep the matched pairs together, personally. $\endgroup$
    – Fomite
    Jan 5, 2016 at 17:43
  • $\begingroup$ @Fomite- actually I'm new to statistics! I want to examine the effect of several biomarkers on insulin resistance among obese and non-obese (cases and controls) and if the association is independent or not so I've conducted multiple linear regression for the full cohort. do i have to conduct logistic regression OR between cases and controls? what do you think! and please if you have a useful reference for conducting a conditional logistic regression using SPSS I would be grateful! thanks again $\endgroup$ Jan 5, 2016 at 18:26

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.