Page 2 of 2
Re: Panel Least Squares estimation method Nonlinear optimization algorithm
Posted: Wed Jul 12, 2017 4:42 pm
by EViews Glenn
The results in EQ01 look fine. You might want to perterb results from those final values to verify stability, but the gradients look good.
Re: Panel Least Squares estimation method Nonlinear optimization algorithm
Posted: Wed Jul 12, 2017 9:17 pm
by acemi
Thank you very much.
Is there method to conduct perturb study in Eviews9 (built-in, add-in, or programming) ?
Re: Panel Least Squares estimation method Nonlinear optimization algorithm
Posted: Thu Jul 13, 2017 9:42 am
by EViews Glenn
Not built-in for NLLS yet. For now, you'll need to write a loop.
Re: Panel Least Squares estimation method Nonlinear optimization algorithm
Posted: Thu Jul 13, 2017 9:53 pm
by acemi
OK thank you.
I would like ask your professional judgement on one issue :
the simulation gave multiple converging results but most of them having different coefficient values. SSR values are also different. I conclude that there are multiple local optima in the search range. Moreover when I rank by SSR values in ascending order, the minimum SSR has some converged coefficient values way off of theoretical expectations. So in the overall I conclude that the results are inconclusive for empirical purposes and a better algorithm should be used in stead.
Best regards