topic Re: How do it interpret treatment effects with ordinal endogenious treatment variable? in SAS Forecasting and Econometrics

How do it interpret treatment effects with ordinal endogenious treatment variable?

Druen — Mon, 17 Jul 2017 12:08:59 GMT

Hi,

I am estimating treatment effects in a model where the treatment variable takes 3 levels (1,2,3) and is endogenous.
My questions are how I can interpret the second-stage coefficients from proc syslin and whether I can compute estimates for a specific value of treatment (say treatment=1).

I run SAS EG 7.1 and have used the following code:

Proc syslin data=sample 2SLS;
endogenous treatment;
instruments Z X;
first_stage: model treatment=Z X;
second_stage: model  Y = treatment+X;
run; quit;

Thanks!

Re: How do it interpret treatment effects with ordinal endogenious treatment variable?

dw_sas — Mon, 17 Jul 2017 15:46:15 GMT

Hi Druen,

This question might be better addressed by contacting Tech Support via the following link:

http://support.sas.com/ctx/supportform/createForm

If you have a reference for the model and the analysis you are trying to perform, then please provide that information as well when you create the Tech Support track.

Best Regards,

Re: How do it interpret treatment effects with ordinal endogenious treatment variable?

Druen — Tue, 18 Jul 2017 14:30:57 GMT

I will try that, thanks!

For future reference, would you mind educating me on why this is a question for Tech Support rather than the forum?

Best ragards

Re: How do it interpret treatment effects with ordinal endogenious treatment variable?

dw_sas — Wed, 19 Jul 2017 14:30:11 GMT

Hi,

Thanks for your follow-up question. I should have provided a reason for suggesting you contact Tech Support in my earlier reply.

Based on the description of your model and data, it was not clear that PROC SYSLIN was the appropriate procedure to use. The SYSLIN procedure is generally used for continuous endogenous variables rather than ordinal data. I thought that a more detailed discussion of your desired model and analysis might be required in order to determine the appropriate procedure(s) to use. That type of detailed discussion is better addressed through Tech Support than through the forum.

I hope that helps!