2-stage Heckman procedure - parameters?

[henribu]

Hello,

I am currently trying to run the two stage heckman method

I am currently rewriting a stata program into SAS and there is a do file containing this method

I started by doing like this website to translate the stata into sas:

https://stats.idre.ucla.edu/sas/code/heckman-selection-model/

It runs ok but later in the stata code they are calling two parameters after this procedure: lambda and sigma

you can learn from these parameters here: https://www.stata.com/manuals13/rheckman.pdf

there are called in stata with e(lambda) and e(sigma)

The problem is the following: I have no idea how to access to these two parameters in SAS

I tried to find it with google without any success

Anyone has an idea?

I am currently using SAS entreprise guide 7.1

Many thanks for your help

[PGStats]

Have you checked the implementation of this method in proc QLIM (part of SAS/ETS)?

https://documentation.sas.com/?docsetId=etsug&docsetTarget=etsug_qlim_details17.htm&docsetVersion=14...

[henribu]

Hi,

Thanks for your answer

Yes I checked this and they are doing exactly what I am doing

The only thing that could be added to what I've mentioned is the heckit option 

But this does not tell me how to get access to the lambda and the sigma

For instance, if I should compute another variable that is depending of the lambda of the estimation,

how I should access to lambda in the code?

Moreover, I tried to replicate the results obtained with stata and they differ

I do not understand why

I ran the two regressions with exactly the same dataset and I obtained different results

proc qlim data=dtest;

model varex = var1 var2 var3/discrete;

model varex2 = var2 var4 var5 select(varex=1);

nloptions maxiter=4000000;

run;

I also tried with the heckit option

proc qlim data=dtest heckit ;

model varex = var1 var2 var3/discrete;

model varex2 = var2 var4 var5 select(varex=1);

nloptions maxiter=4000000;

run;

and I am still obtaining different results than with stata and the following code:

heckman  varex2 = var2 var4 var5   if varex == 1, select(var1 var2 var3)

Many thanks if someone has any idea