Propensity score matching and regression following matching

[Sarah Wan]

Hi

I have never used this technique and have been reading about it but I am still confused. Please could you provide a step by step guide on how to execute.

My treatment indicator t, =1 if received training and =0 if not received training. The sample size is 60 and 522 respectively.

How do I get a one to one match - so match the 60 (in the treated group) with 60 in the control group. Then create a new sample containing the 120 individual observations and then run a regression.

Also, I have seen people talk about x1 x2 etc. being the covariates, are these just all the other z variables in your original regression?

Thanks, Sarah

[Ain Qurat Ul]

I am also searching for this since 3 weeks.

[Sarah Wan]

Ain Qurat Ul - I found this in my research and the response by Ariel really helped me. Hope it helps you too! https://www.researchgate.net/post/Ho...d_using_STATA2

[Ain Qurat Ul]

Sarah Wan you are using stata 14? Cox in my Stata 15 there is no command for psmatch2?

[Joseph Coveney]

in my Stata 15 there is no command for psmatch2

I believe that psmatch2 is user-written, and so won't be present automatically after installation of Stata. You can

search psmatch2

and follow the hyperlinks to a download / installation location.

But you might want to look into what official Stata offers along these lines.

help teffects psmatch

[Sarah Wan]

Yes as Joseph it is user written so you have to install it. You can do so by entering the command ssc install psmatch2. Otherwise you can use teffects which is the newer version actually included into Stata I believe!

A question I have following this is can you use categorical variables when using teffects as it comes up with an error when using psmatch2 (for instance when using a categorical variable with i.VariableName). Do you know if this is possible or maybe there is another way to specify a categorical variable with teffects Joseph Coveney?

[Joseph Coveney]

can you use categorical variables when using teffects

Try it out on the help file's example.

webuse cattaneo2, clear
teffects psmatch (bweight) (mbsmoke i.mmarried c.mage##c.mage i.fbaby c.medu)

[Sarah Wan]

Thanks, thats really helpful!

[rehana anwar]

hey sarah im struggling with the same problem. just wanted to know if u were able to find answer to your query. if yes then i would be really thankful if u share because even after searching the whole youtube and gooogle....im unable to find any guideline on how to run regression on treated and control group. hoping to hear back from you.
P.S. im unable to understand the post of Ariel on research gate... if i assign ids to my treated and control groups, what regression command im gonna use to run test specifically on just those two (treated and control) groups.
regards

[rehana anwar]

furthermore, following the instructions of Ariel leads to the deletion of all the observations from my dataset. need your help