The likelihood function for my model is essentially the exponential distribution with parameter 'lambda', where lambda should be positive by definition.
I want to specify 'lambda' as a linear combination of a set of variables which may cause 'lambda' to be negative for some observations.
In order to keep my 'lambda' always positive, I replace 'lambda' with exp(lambda) in my log likelihood function (As advised in the Parameter Constraints section of the http://www.stata.com/manuals13/rmlexp.pdf).
The program I wrote for this is as follows:
program expo_lf
version 11
args lnfj lambda
quietly replace `lnfj' = ln(exp(`lambda')) - (exp(`lambda')*$ML_y1)
end
My ML command looks like the following:
ml model lf expo_lf (delay= users update_count)
I can generate the coefficients for the for the 2 regressors, but I need to recover the coeffIcients from the impact of the transform I did for lambda.
In cases where constrained parameters are not linear combination of variables, I see that we can recover original parameter using the "nlcom" command, but I am confused in this case.
Your advice is highly appreciated.
Thank You.
I want to specify 'lambda' as a linear combination of a set of variables which may cause 'lambda' to be negative for some observations.
In order to keep my 'lambda' always positive, I replace 'lambda' with exp(lambda) in my log likelihood function (As advised in the Parameter Constraints section of the http://www.stata.com/manuals13/rmlexp.pdf).
The program I wrote for this is as follows:
program expo_lf
version 11
args lnfj lambda
quietly replace `lnfj' = ln(exp(`lambda')) - (exp(`lambda')*$ML_y1)
end
My ML command looks like the following:
ml model lf expo_lf (delay= users update_count)
I can generate the coefficients for the for the 2 regressors, but I need to recover the coeffIcients from the impact of the transform I did for lambda.
In cases where constrained parameters are not linear combination of variables, I see that we can recover original parameter using the "nlcom" command, but I am confused in this case.
Your advice is highly appreciated.
Thank You.
