Hello Statalist,
I have unbalanced panel data for medical providers with 2 million provider-year-level observations. The question that I am trying to answer is how the outcome score of providers varies between program participants vs non-participants. To answer this question, I already used the Inverse Probability of Treatment Weighting (IPTW), but I want to use a different approach using two-level matching.
I want to perform a ps matching after exact matching.
For the exact matching, I use this code:
which gives me a number for a number that I can use for similar providers.
Having this value for exact matching, how can I perform the ps matching? (so it is basically two-level matching)
Also, do you have any suggestion if the number of program participants is only 0.01% of the total sample size (meaning that participation is a rare event)?
Thank you so much for all of your suggestions in advance.
I have unbalanced panel data for medical providers with 2 million provider-year-level observations. The question that I am trying to answer is how the outcome score of providers varies between program participants vs non-participants. To answer this question, I already used the Inverse Probability of Treatment Weighting (IPTW), but I want to use a different approach using two-level matching.
I want to perform a ps matching after exact matching.
For the exact matching, I use this code:
Code:
egen provider_group= group(practice_type specialty)
Having this value for exact matching, how can I perform the ps matching? (so it is basically two-level matching)
Also, do you have any suggestion if the number of program participants is only 0.01% of the total sample size (meaning that participation is a rare event)?
Thank you so much for all of your suggestions in advance.
