1

How can I obtain balance statistics (with MatchBalance of Matching package) for match analysis when I only have the outcome and treatment variable as well as the propensity score but not the different covariates?

This is my match function: 

match1=Match(Y=table$Y,Tr=table$Treat,X=table$PX,replace = T)
summary.Match(match1)
zyurnaidi
  • 2,143
  • 13
  • 14
Nina Kuppetz
  • 11
  • 2
  • 1
    Welcome to Stack Overflow! Please read the info about [how to ask a good question](http://stackoverflow.com/help/how-to-ask) and how to give a [reproducible example](http://stackoverflow.com/questions/5963269). This will make it much easier for others to help you. – zx8754 May 13 '16 at 10:18
  • 1
    You can calculate some balance stats of your PS, but that's it. You need the covariates to perform any other calculations. Just give `MatchBalance` the formula `Y~PS`, filling in the other variables and it will spit out some balance stats related to your PS. Without additional information, such as a sample dataset, I can't test this will run, but I'm pretty sure it will. – lmo May 13 '16 at 11:53
  • @lmo that's correct, but checking for balance on the propensity score alone tells you nothing about the balance on one's covariates. Basically, Nina is screwed without access to covariates, since the very point of PS matching is to create and ensure balance on the covariates. – Noah Jul 07 '16 at 01:03
  • @Noah This steps heavily into CrossValidated territory and I suspect that there are warring parties involved in such things, but Section 14.4 of Imbens and Rubin's 2014 text provides a line of logic supporting the use of the PS for checking the joint covariate balance. This does not allow one to check which covariates are unbalanced, but gives an idea whether there is an overall balance between the groups. – lmo Jul 07 '16 at 12:04
  • Recent authorities, e.g. Austin, and Stuart recommend not assessing balance with propensity scores themselves, but by looking at each covariate balance. King makes a good case for not using propensity scores for matching at all. – Jack Wasey Nov 11 '18 at 01:05

0 Answers0