[Limdep Nlogit List] help request

[Stephen Beville]

Edoardo

Experience has shown me that if the variance is restricted to equal the
mean, the the heterogeneity in the variance estimates can be influenced.
This is something that you must investigate empirically.  I refer you
also to Greene and Hensher 2007.  They use a restricted and unrestricted
triangular distrubtion while controlling for heteroscedasticity.  So the
present both specs.  You might want to do the same?

Good luck,
Stephen Beville
On Sun, Jan 31, 2010 at 4:10 AM, Edoardo Marcucci <emarcucci at uniroma3.it>wrote:

> Stephen, thanks for your suggestions which I will surely try out right
> away.
>
> I also need to impose restrictions on the sign of the coefficients and
> I use U ... ,1 to impose the variance equale to the Beta. do you know
> if this might couse some problems with the estimations..
>
>
> Best regards
>
> Edoardo
> ========================
> Edoardo Marcucci
> DIPES -
> Università degli Studi di Roma Tre
> Via G. Chiabrera, 199 - 00145 ROMA
> tel. ++39  0657335294
> fax. ++39 0657335282
> cell. ++39 3316876905
> mail edoardo.marcucci at uniroma3.it
> ========================
>
>
>
>
>
>
>
>
>
>
> Il giorno 30/gen/10, alle ore 04:04, Stephen Beville ha scritto:
>
> > Edoardo,
> >
> > This problem is easily solved.  Drop the ! and # from your FCN code
> > and if
> > you want all 3 covariates to enter the model (heterogeneity in the
> > variance,
> > do not ascribe any binary coding.  However if you want to restrict
> > some
> > covariates from appearing give these a 0, and in this case the
> > coding must
> > follow |#.
> >
> > For example.
> >
> > For all covariates to enter the model use the following code
> >
> > HFR = INCOME, FAMILY, URBAN
> > FCN = GC(N), TTIME (N)
> >
> >
> > If you don't want family to enter, use:
> >
> > HFR = INCOME, FAMILY, URBAN
> > FCN = GC(N|#101), TTIME (N|#101)
> >
> > BTW-as a credit to the manual, and the writers of the manual, this
> > information is all in there.
> >
> > Cheers,
> > Stephen Beville
> >
> >
> >
> >
> > On Sat, Jan 30, 2010 at 11:22 AM, Edoardo Marcucci <
> emarcucci at uniroma3.it
> > >wrote:
> >
> >> Dear Limdep users,
> >>
> >> We would like to ask a question concerning a problem we have
> >> encountered and are not able to solve. The manual is not very helpful
> >> on this specific point.
> >>
> >> We are trying to estimate  a random parameter model with
> >> heterogeneity
> >> in variance attributable to choice-invariant characteristics.
> >>
> >> the manual (N17-26) shows the code for the heterogeneity both in mean
> >> and variance together
> >>
> >> RPL =  AGE, SEX;
> >> HFR = INCOME, FAMILY, URBAN;
> >> FCN = GC(N), TTIME (N|#01!101)
> >>
> >> WE WOULD LIKE TO STUDY ONLY HETEROGENEITY IN VARIANCE.
> >>
> >>
> >> FOLLOWING THIS EXAMPLE,
> >>
> >> RPL;
> >> HFR = INCOME, FAMILY, URBAN;
> >>
> >> ... WE DO NOT KNOW HOW TO CODE FCN.
> >>
> >> WE HAVE TRIED BOTH:
> >> (WE ACTUALLY USED U INSTEAD OF N)
> >>
> >> 1) FCN = GC(N), TTIME (N|!101)
> >> 2)FCN = GC(N), TTIME (N|#!101)
> >>
> >> but in both cases we could not run the model.
> >>
> >> More in detail:
> >> -  in the first case, the error message was "ERROR 1071: In FCN =
> >> name(type|value), invalid value was found."
> >> -  in the second case, the error message was "ERROR 1071: the #dddd
> >> format is used when ; RPL= list is present"
> >>
> >>
> >> Any help on this point is much appreciated.
> >>
> >> best regards
> >> ========================
> >> Edoardo Marcucci
> >> DIPES -
> >> Università degli Studi di Roma Tre
> >> Via G. Chiabrera, 199 - 00145 ROMA
> >> tel. ++39  0657335294
> >> fax. ++39 0657335282
> >> cell. ++39 3316876905
> >> mail edoardo.marcucci at uniroma3.it
> >> ========================
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >> _______________________________________________
> >> Limdep site list
> >> Limdep at limdep.itls.usyd.edu.au
> >> http://limdep.itls.usyd.edu.au
> >>
> >
> >
> >
> > --
> > Stephen Beville
> >
> > Postdoctoral Fellow
> > Crawford School of Economics and Government
> > The Australian National University
> > Canberra ACT 0200
> > AUSTRALIA
> > O: 61 261254443
> > M: 61 413905168
> > _______________________________________________
> > 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
>



-- 
Stephen Beville

Postdoctoral Fellow
Crawford School of Economics and Government
The Australian National University
Canberra ACT 0200
AUSTRALIA
O: 61 261254443
M: 61 413905168