Estimation of endogenous treatment effects models with -etregress- command

[Johannes Muck]

Dear all,

First of all, I'd like to apologize in advance that this is not a strictly Stata-related topic.

I am currently working on a project with aims at highlighting the ubiquity of self selection problems, i.e. cases in which observation units (e.g. firms or managers) self select into a specific treatment (a specific action).
In our paper, we use a Monte Carlo simulation study to demonstrate how severe the consequences of ignored self selection problems are. To this end, we compare regression results obtained from the -reg- command with those obtained form the newly-introduced Stata 13 command -etregress-.
During the revision process, the referees have also asked us to mention the equivalent commands for etregress in other statistical software packages (e.g. SPSS, R, SAS) to make the paper relevant for a broader audience.
Unfortunately, as a heavy Stata user, I am only familiar with the basics of either R or SPSS. Moreover, Google did not deliver any useful search results which is why I thought about asking you guys.

I know that this forum is about Stata but still I thought maybe some of you might be able to help me out.

Thank you very much in advance.

Best,

Johannes

[wbuchanan]

SPSS almost definitely would not have an equivalent since it isn't an area typically on the radar for their audience. There might be something in R, but the referees are asking a fairly unreasonable request if they want you and/or your team to sift through 7,000 different user contributed programs to determine which if any provide analogous functionality. I work in a shop full of R and SPSS users and have had to have similar conversations with them in regards to causal inference in program evaluation, but the only thing they tend to use are more standard propensity score estimators that don't adjust for endogenous assignment to intervention (so the paper would be incredibly useful for me to further advocate for not using propensity score matching in those cases). If you're posting the code for the simulation publicly and it is reasonably documented/commented I don't see why that couldn't serve the same purpose (e.g., showing the logic/decision rules used to simulate the data followed by the estimation process).

[Johannes Muck]

Dear William,

thank you very much for your reply. I think that it would be no big deal if we weren't able to provide the readers with the corresponding commands in other statistical software packages.

Best,

Johannes