[Limdep Nlogit List] wald test for linear restrictions

[scolombo at ugr.es]

Dear all,

I agree totally on what prof Green say.
Many thanks. I can estimate the model with the restrictions and use a LL
test.

However, I do not understand why the command (; test)does not work even in
a simple conditinal logit model where I use only one ;test option in the
command. This is what is explained in your book applied choice analysis a
primer (page 344-351).
This tool (; test: b(x)-b(y)=0 etc. ) to compare linear specification
versus dummy/effect code specification is very attractive and I think that
is a very common tool used by the limdep users. In my case the command is:

cLOGIT;Lhs=Card; Choices=c1,c2,ref
;model:
U(c1,c2)=ASC+bf*f+bb*b+bcc*cc+btax*cost+br10*r10+br30*r30+br60*r60/
U(ref)=bf*f+bb*b+bcc*cc+btax*cost+br10*r10+br30*r30+br60*r60/
;Test: b(6)-b(7)=0,b(6)-b(8)=0,b(7)-b(8)=0$

I still obtain the same error.
Error   109: ;CLS/RST - error in syntax.  Check specified constraints.


Does somebody know why?

Thanks







> You may have only one ;TEST:... in a command.  However, this device
> cannot be used in this context. You should use a likelihood ratio
> test. You would estimate the model first without the restrictions
> then with the restrictions imposed. To impose the restrictions, you
> must use the ;RST=list form.  However, you do not appear to have
> the right number of coefficients specified in your restrictions. Your
> model has only 4 parameters in the utility functions. That gives 8 for
> the latent class model. There are 8 more in the probability model for
> the classes however, you cannot restrict these.
> /B. Greene
>
> ----- Original Message -----
> From: scolombo at ugr.es
> To: limdep at limdep.itls.usyd.edu.au
> Sent: Friday, March 5, 2010 5:13:09 AM GMT -05:00 Colombia
> Subject: [Limdep Nlogit List] wald test for linear restrictions
>
> Dear all,
>
> I need to know whether I can specify a variable as continuous or as
> discrete in a latent class choice model.
> I used the following estimation code:
>
> NLOGIT;Lhs=Card; Choices=A,B,C
> ;rhs=D,E,F,G
> ;pds=newpnl12 ;maxit=500
> ;Lcm=factor1,factor2,factor3,factor4
> ;pts=2;prob=ClasP; utility=clasUt
> ;Test: b(6)-b(7)=0,b(6)-b(8)=0,b(7)-b(8)=0
> ;Test: b(14)-b(15)=0,b(14)-b(16)=0,b(15)-b(16)=0$
>
> I does not work and the output warnings me that:
>
> Error   109: ;CLS/RST - error in syntax.  Check specified constraints.
>
> I tried to simplify it, even in a simple conditional logit model, (where I
> just use the first set of restrictions)it does not work and repeat the
> same.
>
> Can someboty spot where the error is?
>
> Thanks.
>
>
> _______________________________________________
> Limdep site list
> Limdep at limdep.itls.usyd.edu.au
> http://limdep.itls.usyd.edu.au
> _______________________________________________
> Limdep site list
> Limdep at limdep.itls.usyd.edu.au
> http://limdep.itls.usyd.edu.au
>