Dear Statalists,
I am currently working on a propensity score matching project, looking at the effects of a government programme on labour patterns. Since no control group exists, I am using nearest-neighbour matching with approximately 400 treated and 2000 untreated observations.
After generating the propensity scores using a logit function, I am using the nnmatch command in Stata 12.0 in order to test the effect of the treatment (programme participation) on the outcome (labour hours). This test is successful but I wish to take the analysis one step further by checking gender differences to see if the treatment has different outcomes depending on the gender of the labourer. I have checked several online help sources, manuals and past papers but have been unable to find how to do this. The only option I have so far is to drop observations and run the nnmatch test for each gender seperately, but this is insufficient since it does not allow me to test for statistical significance of the difference between the genders (since I have to split them into two separate samples).
Does anyone have tips on how I can run nnmatch to test the effect of treatment on outcome by gender in a manner that allows for comparison between genders?
Thank you in advance for your help!
Please find my current code below. I am running this on Stata 12.0.
N.B.: The code has been simplified by shortening the list of control variables.
I am currently working on a propensity score matching project, looking at the effects of a government programme on labour patterns. Since no control group exists, I am using nearest-neighbour matching with approximately 400 treated and 2000 untreated observations.
After generating the propensity scores using a logit function, I am using the nnmatch command in Stata 12.0 in order to test the effect of the treatment (programme participation) on the outcome (labour hours). This test is successful but I wish to take the analysis one step further by checking gender differences to see if the treatment has different outcomes depending on the gender of the labourer. I have checked several online help sources, manuals and past papers but have been unable to find how to do this. The only option I have so far is to drop observations and run the nnmatch test for each gender seperately, but this is insufficient since it does not allow me to test for statistical significance of the difference between the genders (since I have to split them into two separate samples).
Does anyone have tips on how I can run nnmatch to test the effect of treatment on outcome by gender in a manner that allows for comparison between genders?
Thank you in advance for your help!
Please find my current code below. I am running this on Stata 12.0.
N.B.: The code has been simplified by shortening the list of control variables.
Code:
*running logit regression of control variables on programme participation logit programme age city land house educ health *predicting propensity scores predict pscore_programme *saving dataset for future use of propensity scores save programmedata, replace *running nnmatch for men keep if gender == 1 nnmatch labour programme age city land house educ health, m(1) nnmatch labour programme age city land house educ health, m(2) *running nnmatch for women use programmedata, clear keep if gender == 0 nnmatch labour programme age city land house educ health, m(1) nnmatch labour programme age city land house educ health, m(2)
