I have a dataset of two patient and one healthy control group which I would like to compare (using R) with respect to a continuous outcome variable (each subjects is measured once). However the groups differ in age. In the context of the dataset the age effect is not of interest. My questions are:

  • What is the best way to account for the potentially confouding effect of age?

  • From what I know one could do this by either adding it as a covariate in the linear model or matching the samples with respect to age via usage of the propensity score. Which method is recommended in which context?

  • Is there an function to select subsamples of each group while matching groups with respect to age?


Michael Posner wrote an interesting paper comparing propensity score matching with regression adjustment and a third technique using instrumental variables. The propensity score matching approach worked impressively in his examples. I will try to get you a link to the paper.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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