[Limdep Nlogit List] probit vs. semiparametric model

William Greene wgreene at stern.nyu.edu
Wed Sep 30 20:15:34 EST 2009

(1) There is no test statistic for "testing" a parametric specification
against a semiparametric one. They are not estimating the same thing and
they are not maximizing the same criterion function.  The models are not
nested in any sense save for the "space" of models in the parametric 
case is narrower than that for the semiparametric one.
(2) You can provide your own starting values for the Klein and Spady model.
But, to do so, make sure you put the constant term, ONE, last in your
RHS list. Then, also assume that the starting value for the constant term
must be 0.0 and the starting value for the second to last coefficient, 
just before the constant term, must be 1.0.
/Bill Greene

----- Original Message -----
From: "Jadwiga Kostrzewska" <dwakotki at wp.pl>
To: "Limdep and Nlogit Mailing List" <limdep at limdep.itls.usyd.edu.au>
Sent: Wednesday, September 30, 2009 3:43:46 AM GMT -05:00 Columbia
Subject: [Limdep Nlogit List] probit vs. semiparametric model

Hello Everybody once again :)
I ask too quickly my previous question about Chi2 tests, and I know the 
answer now.
But I do not know how to test the probit model vs. its semi-parametric 
specification received by Klein’s and Spady’s method.
Any ideas are welcome.

I have still a question about adding ; Start=... to give my own start 
values for SEMI?

Jadwiga Kostrzewska

Tysiące ofert mieszkań!
Sprawdź nowe ceny.

Limdep site list
Limdep at limdep.itls.usyd.edu.au

More information about the Limdep mailing list