topic Re: nonlinear regression in SAS Forecasting and Econometrics

nonlinear regression

sasphd — Mon, 04 May 2015 15:01:23 GMT

I want to estimate this equation y=a+b1*x+b2*x**(2) using gmm

When I write this program it does not work

proc model data= bb.mstylevw;

parms a b1 b2 ;

y=a+b1*x+b2 *x**(2);

fit y / gmm;

run;

quit;

when I remove Gmm, it works and it gives me regression using OLS

proc model data= bb.mstylevw;

parms a b1 b2 ;

y=a+b1*x+b2 *x**(2);

fit y / ;

run;

quit;

Re: nonlinear regression

SteveDenham — Fri, 08 May 2015 15:36:09 GMT

I am not an expert on PROC MODEL, but I thought that gmm would require an INSTRUMENTS statement, which here would be pretty simple. If you explicitly specify the moments, then it might be an intercept only situation.

For examples, see 19.18 Duration Data Model with Unobserved Heterogeneity in the SAS/ETS13.2 documentation.

Steve Denham