Dear statalist,
I want to use -psestimate- (from SSC) to select a linear function of covariates to include in the estimation function of the propensity score. At first, I put around 20 variables in totry(), it took very long for stata to run. Then I reduced the list to 10 variables, and then to 5. Every time the final model stata gives me is all variables I put in totry(), and if I use those variables to run -psmatch2-, the t-test between matched treatment and control groups are still significant for most of the variables.
I wonder what I did wrong, or is that the list of variables I choose cannot achieve insignificant difference between matched treatment and control groups post-match?
My code
and I get
Thanks!
I want to use -psestimate- (from SSC) to select a linear function of covariates to include in the estimation function of the propensity score. At first, I put around 20 variables in totry(), it took very long for stata to run. Then I reduced the list to 10 variables, and then to 5. Every time the final model stata gives me is all variables I put in totry(), and if I use those variables to run -psmatch2-, the t-test between matched treatment and control groups are still significant for most of the variables.
I wonder what I did wrong, or is that the list of variables I choose cannot achieve insignificant difference between matched treatment and control groups post-match?
My code
Code:
psestimate treat, totry(asset score SOE marketcap leverage) noquad
Code:
Selecting first order covariates... (15) ----+--- 1 ---+--- 2 ---+--- 3 ---+--- 4 ---+--- 5 ....s...s..s.ss. Selected first order covariates are: asset score SOE marketcap leverage Final model is: asset score SOE marketcap leverage
