selmlog "invalid syntax" error

[Juan Felipe Contreras]

Dear Stata programmers,

I am intending to estimate an ATE using a multinomial logit selection correction regression model with the selmlog command. However, I am receiving the "invalid syntax" error, which I do not understand since I think I have followed the instructions to how to perform this estimation. My code is as follow:

selmlog y x, select(cat=z) dmf(1)

where y is a continuous variable that has been standardized; x is a discothomous variable; cat is a four level categorical variable representing the multiple counterfactuals; and z is an explanatory variable for the multiple choices of the cat variable.

Any help on understanding what I am doing wrong would be appreciated.


Kind regards,

Juan

[William Lisowski]

As I write this, selmlog is neither a Stata built-in command nor a package that can be found using search selmlog. So by not providing information on where to obtain this community-contributed command you are limiting your possible respondents to those who have found and used the command themselves, excluding those who while unfamiliar with selmlog are willing to invest some time in your problem.

With that said, perhaps in the 16 intervening years since selmlog was created, Stata's built-in capabilities have expanded to supplant the need for selmlog. The sem and gsem commands in particular come to mind.

[Juan Felipe Contreras]

William, you are right, the selmlog command is not a built-in nor a package, rather an external dependency that can be stored locally via an .ado file. The documentation is available in selmlog13
(parisschoolofeconomics.com).

The instructions regarding syntaxis is www.parisschoolofeconomics.com/gurgand-marc/selmlog/selmloghelp.pdf

Thanks.

[William Lisowski]

I made up an example and ran selmlog; I did not encounter the syntax error. For further investigation, I think you will have to create an example that causes the problem on a smallish dataset that you can easily share using the output of the dataex command.