Hi Guys,
Recently, I have been using STATA to do a propensity score matching. My codes are very simple:
psmatch2 y x i.country_FE i.industry_FE i.time_FE, noreplacement caliper(.01) neighbor(1)
So, basically, I matched every firm with y=1 to a firm with y=0 based on x (a series of firm characteristics), country-fixed effects, industry-fixed effects, and time-fixed effects.
After a lot of fixed effects predicting failure perfectly and being dropped, I got the PSM outcomes. However, I tried typing "unique _id" and found that they are not unique! Besides, although I had specified "noreplacement," there are observations with _weight>1, meaning they are used more than once. The results are against the options I specified (if I have understood them correctly).
Has anyone encountered a similar situation before? By the way, I am using a very large dataset (almost 80GB). If I psmatch2 a subsample of the dataset (approximately 10GB), the outcomes are normal. _id are unique, all _weight are equal to one. Is it an issue because of the size of the data?
It is my first time posting here, so I am not sure if I have described the issue correctly. Many thanks in advance for your understanding and help.
Best,
Eugene
Recently, I have been using STATA to do a propensity score matching. My codes are very simple:
psmatch2 y x i.country_FE i.industry_FE i.time_FE, noreplacement caliper(.01) neighbor(1)
So, basically, I matched every firm with y=1 to a firm with y=0 based on x (a series of firm characteristics), country-fixed effects, industry-fixed effects, and time-fixed effects.
After a lot of fixed effects predicting failure perfectly and being dropped, I got the PSM outcomes. However, I tried typing "unique _id" and found that they are not unique! Besides, although I had specified "noreplacement," there are observations with _weight>1, meaning they are used more than once. The results are against the options I specified (if I have understood them correctly).
Has anyone encountered a similar situation before? By the way, I am using a very large dataset (almost 80GB). If I psmatch2 a subsample of the dataset (approximately 10GB), the outcomes are normal. _id are unique, all _weight are equal to one. Is it an issue because of the size of the data?
It is my first time posting here, so I am not sure if I have described the issue correctly. Many thanks in advance for your understanding and help.
Best,
Eugene
