In NLogit software, when constructing a random parameter model（or MNL model）, each utility function initially contains all types of variables. How to gradually eliminate unnecessary (or insignificant) ?
There are two specific situations: (1) when all variables are included, error 803 is reported;
(2) Many coefficients appear as fixed parameters, and the estimated parameter for the last utility function is all fixed parameters.
(Note:I also tried BFGS, but it couldn't even come up with this table)
- Badges
- Science topic
- Similar topics
- Geoscience
- Asia
- India
More Jhon Wenver's questions See All
Similar questions and discussions