Hi everyone,
Unfortunately due to a thesis supervisor mismatch she is not able to help me out, so I hope you can. I am investigating the effect of hallucinogens use on anxiety level. My panel dataset contains data of 14,693 individuals of which 106 have used hallucs in one or more years. In total there are 13 annual waves, but not everyone has filled in every questionnaire so there's missing data. For hallucs use I have both a binary and a continuous variable and anxiety level is measured on scale of 1-6.
I want to perform Propensity Score Matching (PSM) based on 2 continuous, 5 categorial and 5 binary variables. I want to match individuals based on one year, so not based on all 13 waves (because then no matches are found). So basically the 13 observations of a single individual are seen as identical individuals.
However, I know three different matching commands: psmatch2, teffects nnmatch and teffects psmatch.
My main question is: Which command should I use?
Ilse
Unfortunately due to a thesis supervisor mismatch she is not able to help me out, so I hope you can. I am investigating the effect of hallucinogens use on anxiety level. My panel dataset contains data of 14,693 individuals of which 106 have used hallucs in one or more years. In total there are 13 annual waves, but not everyone has filled in every questionnaire so there's missing data. For hallucs use I have both a binary and a continuous variable and anxiety level is measured on scale of 1-6.
I want to perform Propensity Score Matching (PSM) based on 2 continuous, 5 categorial and 5 binary variables. I want to match individuals based on one year, so not based on all 13 waves (because then no matches are found). So basically the 13 observations of a single individual are seen as identical individuals.
However, I know three different matching commands: psmatch2, teffects nnmatch and teffects psmatch.
My main question is: Which command should I use?
- Should I look at ATE or ATET? teffects nnmatch provides both, teffects psmatch gives the error message "treatment overlap assumption has been violated" when looking for ATE.
- psmatch2 gives the opportunity to create a separate dataset with only the treated and their matches, by using
Can/should I perform a linear regression or Difference-in-Difference on this sub-sample? Or is the ATT result that psmatch2 provides what I am looking for?Code:
drop if _weight == .
Ilse
