I am trying to estimate a random parameter linear regression models in LIMDEP (Hildreth, Houck and Swamy's Random Coeffficents Model). Since my data is x-sectional, therefore, I am not using Pds command (being x-sectional data), my commands are as follows:
regress;lhs=......
;rhs= one,....................
;Fcn= .......................
;RPM
;Pts=50
;Halton$
LIMDEP works in two step: 1. In step-one LIMDEP use OLS and comes up with starting values for ramdom parameter model and outputs an initial model. My initial model(starting value) as R-Squred of 0.5. It also gives log-likelihood for starating model which in my case is -6600. 2. Instep-2 LIMDEP uses results from OLS model to estimate random parameter regression models. In this step LIMDEP uses maximum likelihood estimates, therefore it outputs log-likelihood at zero (-6900) and at convergence is (-6700).No more R-Squred values in step-2 as LIMDEP is using different technique. 3. I am having issue of the overall fit of the model. If I try to calculate Rouhe-Squred, it is less than 0.02 which is not very appealing. On the other hand I see that starting models used by LIMDEP to estimate random parameter linear regrssion models has an R-Squred of 0.5 which is good enough. In seceond step when LIMDEP gives final results it does not show R-Squred, instead it gives log-likelihood at zero (-6900) and at convergence (-6700).I have 7 total significant variable (four random parameter) all having T-value >2 and P-value<0.001).
My question is how should I report my overall fit and is my model OK.
