Re: st: nlogitrum and nlogit: unbalanced data

[John Fulton <fulton@ssc.wisc.edu>]

> Richard Gates suggested modifying nlogitrum to remove a trap.  
> I tried it and it didn't work.
Hi Statalist.
I followed Richard's suggestion to modify -nlogitrum- to 
"-mynlogitrum-."  "-mynlogitrum-" returns the correct tree, then gives 
an error: "weights must be the same for all observations in a group."

I assume I'm running afoul of another kind of check for balanced data, 
or possibly a check for degenerate nests.

I didn't need to do it, but I egened min and max weights by group, and 
selected cases where they were unequal.  There were none.

Again, the -nlogit- runs fine.

Thanks,
John.


Richard Gates wrote:

> John Fulton is inquiring about the "unbalanced data" error message from
> -nlogitrum-.
>
> There is an error trap in -nlogitrum- to check that the alternatives for each
> case (individual) are balanced.  This error trap is unnecessary.  The code in
> -nlogitrum- is based on the Stata 7 release of -nlogit-.  Since then we 
> (at StataCorp) have determined the restriction that each case must have the
> same set of alternatives is unnecessary and we have removed it from -nlogit-.
>
> John, if you are feeling up to a bit of programming you can copy nlogitrum.ado
> to your working directory and modify it.  You can find where -nlogitrum-
> resides using the stata command
>
> . which nlogitrum
>
> It is probably in your personal ado directory. 
> You will see in the ado code a call to the local program -ChkDta- on line 78.
> Comment out this line.  
> Your ado path will have the working directory ahead of the personal directory.
> To check your ado path use
>
> . adopath
>
> Clear the original version of -nlogitrum- from memory using
>
> . program drop _all
>
> then try running your model again.
>
> Alternatively, you could copy nlogitrum.ado to mynlogitrum.ado and
> change the program name within the file to mynlogitrum.ado.  
> Be sure to constrain the inclusive-valued (dissimularity) parameters to 1
> for the degenerate nests in the RUM model otherwise you will not get 
> convergence.  I do not believe the IV parameters exist for the RUM
> model in this case.  They are identifiable in the Nonnormalized model
> if you have an alternative specific variable (maybe var1 in your model
> specification). 
>
> Best Regards
>
> -Rich
> rgates@stata.com
> *
> *   For searches and help try:
> *   http://www.stata.com/support/faqs/res/findit.html
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
>   
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/