[Limdep Nlogit List] negative IIA test statistic

William Greene wgreene at stern.nyu.edu
Sun Jun 24 01:34:12 EST 2007


Aleks. You have to strip off the last two rows and columns.  You can 
only compare coefficients that are estimated under both scenarios.
You don't have a comparison for the last two.
/Bill

************************************************
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: Wednesday, June 20, 2007 4:42 pm
Subject: [Limdep Nlogit List] negative IIA test statistic

> Bill,
> Vn is not positive definite, indeed.
> But no wonder there are negative diagonal elements - it comes from the
> substraction Vr-Vu.
> the code, as in the book, for my 19 parameters, 6 attributes model is:
> MATRIX; J2=[
> 1,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0/
> 0,1,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0/
> 0,0,0,1,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0/
> 0,0,0,0,1,0,0,0,0,0, 0,0,0,0,0,0,0,0,0/
> 0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0/
> 0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0] $
> 
> MATRIX ; Br=J2*B ; Vr=J2*VARB*J2' $
> 
> ? so in Vr the last two rows and two columns are all zeros, while 
> in Vu
> there are values >0 (there was no elements set to 0 for the 
> unrestrictedmodel)
> 
> MATRIX ; bn= Bu-Br ; vn= Vr-Vu $
> 
> ? as a consequence, in vn the two last diagonal elements are negative.
> 
> MATRIX ; vninv=[vn]$
> MATRIX ; list ; q=bn'*vninv*bn$
> 
> Sorry but I can't find a mistake there, it seems that in the 
> example in
> Applied Choice Analysis (p. 600) there is also a row and a column 
> of zeros
> in Vr, so after substracting Vu (which is not shown) one diagonal 
> element of
> vn would be negative...
> Where am I wrong then?
> 
> Aleks
> _______________________________________________
> Limdep site list
> Limdep at limdep.itls.usyd.edu.au
> http://limdep.itls.usyd.edu.au
> 




More information about the Limdep mailing list