nlogit / asclogit replicate choice-alternatives per case

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

nlogit / asclogit replicate choice-alternatives per case

Nils Wlömert
Dear listers,

my choice desingn includes multiple consecutive choice decisions per  
respondent (8 choices á 4 alternatives) and in some cases an  
alternative occurs twice per case (in different choice decisions).

This seems to cause problems estimating asclogit and nlogit and the  
following error messages occur:

nlogit: "note: 2324 cases (74368 obs) dropped due to no positive  
outcome or multiple positive outcomes per case"
asclogit: "variable alternative has replicate levels for one or more  
cases; this is not allowed"

(How) Is it possible to estimate asclogit and nlogit with multiple  
consecutive choices and replicate alternatives (in different choice  
decisions) per case? I have tried estimating both models with only one  
choice decision per case (4 alternatives each) which works fine.

Also, when estimating asclogit with effects-type coding the model does  
not converge due to collinearity (3 out of 26 parameters are dropped).
asclogit: "note: model has collinear variables; convergence may not be  
achieved"
What would be the best way to deal with collinarity in this case? The  
model does not converge when the parameters are kept.

Many thanks!

Nils.





*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
Reply | Threaded
Open this post in threaded view
|

Re: nlogit / asclogit replicate choice-alternatives per case

jdai
This post has NOT been accepted by the mailing list yet.
I am facing the same problem. Have you already figured out how to solve it?
Reply | Threaded
Open this post in threaded view
|

Re: nlogit / asclogit replicate choice-alternatives per case

syntegrity
This post has NOT been accepted by the mailing list yet.
I've been looking around the web for an answer to this problem myself, and divined the answer by scrutinizing the restaurant example data set in the nlogit helpfile and comparing it with my own data set. Hopefully I can save you having to go to the same trouble!

The problem is probably in your nlogittree structure. You probably have the wrong variable included for one of the levels in your tree diagram. I'm guessing its the lowest level, as that was my error. The lowest level of your branch structure (i.e. the first variable you specify after the nlogittree command) should be the list of alternatives. That means that it lists a different alternative in every row, repeating the same list of alternatives over and over again for each case in your data set.
Example Data Set
In the example above, it is the variable restaurant that is the lowest level of the branch structure.

Hope that helps!

Best,
Chris Congleton