[Limdep Nlogit List] The partial effects in non-linear utilityfunctions in the Ordered Choice models

[Andreas Drichoutis]

Simona, what is the formula you use for the overall partial effect? From my
understanding it should not make a difference and you should only observe
one sign change, unless I'm missing something.
In the case where you only included age (and not age squared) the partial
effect would be something like this:
PE=b1*(f(MU1-b'X)-f(MU2-b'X))
Now if we assume that b2 is the coefficient of age squared you should
compute:
PE=(b1+2*b2*Age)*(f(MU1-b'X)-f(MU2-b'X))
Are you computing these at variable means?

Kind regards,
Andreas

-----Original Message-----
From: limdep-bounces at limdep.itls.usyd.edu.au
[mailto:limdep-bounces at limdep.itls.usyd.edu.au] On Behalf Of Simona Rasciute
Sent: Thursday, June 25, 2009 11:45 AM
To: limdep at limdep.itls.usyd.edu.au
Subject: [Limdep Nlogit List] The partial effects in non-linear
utilityfunctions in the Ordered Choice models


Dear All, 

I am trying to compute partial effects in the Ordered Choice model when the
utility function is non-linear, following Greene and Hensher (2009)
"Modeling Ordered Choices". For example, the list of explanatory variables
includes not only the age of the respondent but also age squared. All
estimated partial effects are subject to the single crossing feature (one
sign change). Does the single crossing feature have to also apply to the
computed partial effects when non-linear variables are included? When I
compute the overall partial effect for the age variable, there are two sign
changes. 

Thanks a lot for your help.

Kind Regards,
Dr Simona Rasciute
Research Associate

Department of Economics
Loughborough University
Loughborough
Leicestershire
LE11 3TU
Tel: +44(0)1509 223622



_______________________________________________
Limdep site list
Limdep at limdep.itls.usyd.edu.au
http://limdep.itls.usyd.edu.au