PSM - matchit variable questions

Asked Sep 20 '22 at 16:57

Active Sep 20 '22 at 17:43

Viewed 17 times

I am estimating a propensity score for use in a logistic regression. For this, I am using the matchit command. I have a couple questions on variable inclusion when running PSM:

Is there a limit/best practice on how many variables I should use in the match, similar to the one in ten rule for predictor parameters?
Is there a problem associated with using a categorical variable?

asked Sep 20 '22 at 16:57

tchoup

There are many reasons not to use matching, detailed here. There are also many reasons for not doing propensity score analysis as discussed there. – Frank Harrell Sep 20 '22 at 17:45
3

Does this answer your question? Is it possible of overfit using Propensity score matching with the MatchIt R package? Although the title of the question seems different from yours, I think that the answer on that page covers the issues that you raise (and some that you don't but you should be aware of) pretty well. If there are still unresolved questions, please edit this question to be specific about them. – EdM Sep 20 '22 at 20:20
2

Agree with @EdM. This answer may help with respect to including categorical covariates in matching. The answer: there isn't a problem, and why would there be? – Noah Sep 20 '22 at 20:54

PSM - matchit variable questions

0 Answers0