John Fulton is trying to run a (degenerate) nested logit model. He
states below:
> Here's the form of my data with a hypothetical example: a study of
> choices between cake and various kinds of pie. Please ignore the
> independent variables - they're completely made up and I include them
> only to clarify the syntax.
>
> The nlogittree output would be:
> top---->bottom
> dessert type pie
> __________________________________
> cake cake chocolate
> pie fruit_pie cherry
> peach
> cream_pie banana
> coconut
> the model would be:
> nlogit chosen (pie=calories) (type=pie_rating) (dessert=freshness),
> group(person_id)
>
> As you see, the "cake" nest is degenerate. My model kicks out an error:
> "Equation/parameter /cake/ multiply defined"
This is a coeffcient name space problem. The cake level is defined
twice in the labels for the dessert and type variables. Moreover,
the label pie in the dessert variable will confict with the variable
named pie. Renaming the labels for dessert to _cake and _pie and I
believe the estimation will proceed. The coefficents will then be
named
regression coeffcients
pie:calories
type:pie_rating
dessert:freshness
Inclusive value parameters
_cake:_cons
_pie:_cons
cake:_cons
fruit_pie:_cons
cream_pie:_cons
John may also need to constraint the [_cake]_cons and
[cake]_cons estimates to 1.
-Rich
[email protected]
*
* 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/
