The Basics of PSM - Propensity Score Matching

[Jess Lloyd]

I want to run a PSM. My understanding of this technique is that you must sort your data into a control and treatment group. In my case, areas that have been, and have not been effected by conflict. So far I have generated a binary variable for whether of not conflict occurred using the code

" gen highconflict = 0 /// replace highconflict = 1 if conflict>10"

However, I have been told that when running a PSM, there is a different way in which you must create your treatment and control group. Does anyone know the process?

Thanks

[Mari Meir]

Hello Jess,

As I am not sure what you mean by "different way in which you must create your treatment and control group", I can only recommend readings.

If you have already read the seminal paper Rosenbaum and Rubin's seminal paper (1983 The central role of the propensity score in observational studies for causal effects) , you can check papers that cite it and that use PSM so you understand how they do it.

https://scholar.google.fr/scholar?hl...c=&q=&scisbd=1

There are videos that show how to implement it, they can be helpful.

And you can find some guidance here: Caliend and Kopeinig 2008 SOME PRACTICAL GUIDANCE FOR THE IMPLEMENTATION OF PROPENSITY SCORE MATCHING

For further questions, make sure to read the FAQ of StataForum so you can get more meaningful and helpful comments.

Best,

MM

[Phil Bromiley]

Stata has an entire manual on treatment effects that includes propensity score matching. Naturally, if it does what you want, using a pre-written procedure beats trying to program it yourself.