[Limdep Nlogit List] negative IIA test statistic

[William Greene]

Aleks. The problem is the matrix you inverted to get VNINV. It is not positive
definite, so the inversion returned something unusable.  Take a look at the
matrix (before inversion).  Are there rows and columns of zeros?  Negative diagonal
elements, etc.  If so, go back and examine how the matrix was created.
/Bill Greene

************************************************
Professor William Greene
Department of Economics
Stern School of Business
New York University
44 West 4th St., Rm. 7-78
New York, NY   10012
Ph. 212.998.0876
Fax. 212.995.4218
URL. http://www.stern.nyu.edu/~wgreene
Email. wgreene at stern.nyu.edu
************************************************

----- Original Message -----
From: A N <aleks.nnn at gmail.com>
Date: Monday, June 18, 2007 2:46 pm
Subject: [Limdep Nlogit List] negative IIA test statistic

> Dear all,
> 
> When performing IIA test manually as described in Applied Choice 
> Analysis, I
> get a negative test statistic :
> 
> MATRIX ; list ; q=bn'*vninv*bn$
> 
> Matrix Q        has  1 rows and  1 columns.
>               1
>        +--------------
>       1|  -25.75191
> --> Calc ; list ; p=1-chi(q,6) $
>  Error   471: Cannot compute CHI; parameter 1 is < = 0
>  Error   116: CALC - Unable to compute result. Check earlier message
> 
> The Calc won't work, but there is no reason to reject the IIA 
> assumptionbecause the critical value for the test is >0 so any 
> negative q is ok., am I
> right?
> How can I calculate the p-value, then?
> 
> Or is negative q an incorrect output?
> 
> Hope it's not a stupid question, I'm still learning to use nlogit ...
> Comments will be much appreciated.
> 
> Aleks
> _______________________________________________
> Limdep site list
> Limdep at limdep.itls.usyd.edu.au
> http://limdep.itls.usyd.edu.au
>