[Limdep Nlogit List] Maximize command with Panel data

Ashutosh Patil patil_ashutosh at yahoo.com
Tue Jun 17 03:24:03 EST 2008


Dear Professor Greene and other Limdep Gurus:
 
When specifying random parameter models, it seems that Limdep permits the analyst to specify only one common RPM specification for all the parameters that have been specified as random. In other words, Limdep does not seem to provide a way to specify a different RPM specification for different random parameters. Or, am I missing something? 
 
One method of overcoming this, may be to write ones own MLE function in Limdep (though programming such a complex random-parameters likelihood function will be really complex and frought with dangers such as programming an incorrect function, erroneously). But, continuing on that thought, I wanted to get some inputs on specifying MLE function for panel data models, in Limdep. A panel data likelihood function will require the analyst to calculate the likelihood for each group and then sum that group-level likelihood across all the n groups in the data, and then optimize that final sum. However, Limdep’s maximize command seems to automatically apply the same likelihood function to all the observations (within and across all the groups). Since Limdep provides random effects on such an extensive selection of models, Limdep code is obviously doing this internally. I tried to do a lot of thinking about how I could do that (I thought in terms of using the
 “Sample” command repeatedly to ensure that I only picked one group at a time etc.). But that does not seem to be the correct way to proceed. Please advice, on a more elegant way to think about how I may proceed on enabling panel data specification within the maximize command. Once I know how to get that done, I will be able to write the maximize function, on my own (hopefully, without coming errors). 
 
One final question, can Limdep’s maximize command be used for multiple-equation models, with multiple endogeneous variables per equation and mulitivariate normal distribution of the disturbance term? 
 
As always, I am truly grateful for your advice. 
 
Kind Regards,
Ashutosh Patil (doctoral student – Georgia Tech)


      



More information about the Limdep mailing list