Hi Stata users,

I am using Stata 14 to run a random effects model for a panel of 21 countries during the period (2000-2017) with 378 observations. I am trying to determine the best functional form for my model, so I am comparing between a Cobb-Douglas function and a Translog function. I have estimated the model using maximum likelihood estimation for both functional forms. When I compare between the 2 models using lrtest (as shown below), the LR Chi2 is negative. Is that possible? Would then the interpretation be that I cannot reject the null hypothesis (since the p-value is greater than 0.05) and hence the Cobb-Douglas form is preferred?

Thanks.

Code:
 
lrtest retrans recb

Likelihood-ratio test                                             LR chi2(27) =   -197.68
(Assumption: recb nested in retrans)                  Prob > chi2 =    1.0000