I'm trying to match consecutive patients by fitting a Propensity Score. The treatment has 3 levels (Controls, Treat_1, Treat_2). The "MatcheIt" package, is designed to work only with 2 levels treatment. I tried with the "Twang" package, however it seems to be designed for Propensity adjusted analysis, not for matching. My outcome of interest is time to event and I need to plot survival curves and I think that PS-matching may be superior to PS-adj.

  • $\begingroup$ Hi Gabriele, nice you are joining CV. Perhaps this post will be useful: stats.stackexchange.com/questions/202698/comparing-two-or-more-treatments-with-inverse-probablity-of-treatment-weighting/203771#203771 $\endgroup$ – Joe_74 Apr 1 '16 at 13:28
  • $\begingroup$ Gary King has papers criticizing using PS for matching and they are pretty compelling. He recommends using a mahalanobis distance function instead. See this ungated paper Why Propensity Scores Should Not Be Used For Matching here... gking.harvard.edu/files/gking/files/psnot.pdf?m=1456683191 And there's an even clearer explanation in his ICM webinar last fall ... thepoliticalmethodologist.com/2015/09/02/… $\endgroup$ – Mike Hunter Apr 1 '16 at 13:34

Your Answer

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

Browse other questions tagged or ask your own question.