topic Re: Instrumental variable analysis - 2SLS regression and clustering in SAS Forecasting and Econometrics

Instrumental variable analysis - 2SLS regression and clustering

Shad — Wed, 12 Jul 2023 17:16:35 GMT

Hello,

I'm interested in performing a two-stage least-squares regression using an instrumental variable to compute estimated values of the predictor (in the first stage), and then uses those computed values to estimate a linear regression model of the dependent variable (the second stage). From what I've gathered this can be achieved using PROC SYSLIN. However, I'm working with multilevel data and would like to cluster patients within hospitals, I have not seen any documentation regarding clustering/multilevel/hierarchical data in proc syslin, perhaps I've missed it or is there another procedure that better supports the type of analysis?

/* 2SLS estimation */
proc syslin data = analytic_iv 2sls; 
	endogenous continuous_opioid_sum; 
	instruments move_avg_1year; 
	model log_Infla_Est_Cost = continuous_opioid_sum;
run;

Best regards,

Re: Instrumental variable analysis - 2SLS regression and clustering

sbxkoenk — Thu, 13 Jul 2023 20:51:32 GMT

Hello,

I have moved your post to the "SAS Forecasting and Econometrics"-board.

I have not heard before about a multi-level 2SLS regression (using instrumental variables (IV)).

With PROC MODEL (available in SAS/ETS, just like PROC SYSLIN), you can do:
Heteroscedastic Two-Stage Least Squares Regression.
https://support.sas.com/rnd/app/ets/examples/h2sls/index.htm

https://go.documentation.sas.com/doc/en/pgmsascdc/v_040/etsug/etsug_model_toc.htm

Not sure if that will bring you any further.

BR, Koen

Re: Instrumental variable analysis - 2SLS regression and clustering

ChanceTGardener — Tue, 18 Jul 2023 00:02:32 GMT

The CPANEL procedure in SAS Econometrics supports 2SLS on panel data

https://support.sas.com/kb/67/951.html

SAS Help Center: ENDOGENOUS Statement