[Limdep Nlogit List] Multinomial Multiperiod Probit Model

William Greene wgreene at stern.nyu.edu
Mon May 22 02:49:16 AEST 2017

Melanie.  NWAVES should give the number of choice situations per household,
not the number
of rows of data.  If you have 3 choices in every choice situation and 4
choice situations for each
household, NWAVES would equal 4, not 12, as you noted.  With varying
numbers of choice
situations, nwaves is going to take that value for each row for each
household.  For the households
with 4 waves, NWAVES will equal 4 for the 12 rows.  If the household was
observed 5 times,
NWAVES would equal 5 for 15 rows.
That said, I doubt the multinomial probit will work well with varying
numbers of
choice situations per household, because the full covariance matrix, which
is NWAVES*3 by
NVAVES*3 is being estimated, so when this has varying dimensions, the log
likelihood is going
to become ambiguous to compute.  It will work fine for mixed logit models
and LC models.  But,
for the MNP model, the covariance matrix is a fixed entity with fixed
dimensions.  You might make
some initial progress by selecting (for purpose of getting started)
households that all have the same
number of waves.  For more general purposes, I suspect the mixed logit, or
the error components
logit model will work out better.
Bill Greene

On Fri, May 19, 2017 at 10:33 AM, Melanie Velarde <
melanie.velarde at uni-osnabrueck.de> wrote:

> Hello everyone,
> I'm using Nlogit for the very first time and I'm trying to estimate a
> multiperiod multinomial probit model on the basis of an unbalanced panel
> (household survey).
> To keep things simple, I started with the following specification (fyi I'm
> not working with any choice specific attributes)
> ;Lhs = CHOICE
> ;Choices = A, B, C
> ;Model:
>             U(B)= BALPHA + BDCE*DCE /
>             U(C)= CALPHA + CDCE*DCE $
> and Nlogit started to "work" without any complaints. However, as soon as I
> add
> ;Pds = NWAVES
> to take account of the fact that I'm using an unbalanced panel, I get the
> following error message
> Error   1097: Unusable LHS variable in choice model setup
> I've read all the information about preparing and working with panel data,
> which is available in all three guides provided when purchasing Nlogit.
> I particularly made sure that my variable NWAVES reflects not the number
> of rows per household (I have 3 rows per survey wave and household, because
> of the 3 identical choices available every wave), but rather the total
> number of waves that a household was surveyed.
> I have no clue what I'm doing wrong! Any ideas?
> Thanks a lot in advance for any suggestions!!!
> Best regards,
> Melanie Velarde
> Wissenschaftliche Mitarbeiterin
> Fachgebiet Ökonometrie und Statistik
> Fachbereich Wirtschaftswissenschaften
> Universität Osnabrück
> Rolandstr. 8
> 49069 Osnabrück
> Tel.: + 49 541 9692756
> E-Mail: melanie.velarde at uni-osnabrueck.de<mailto:melanie.
> velarde at uni-osnabrueck.de>
> _______________________________________________
> Limdep site list
> Limdep at limdep.itls.usyd.edu.au
> http://limdep.itls.usyd.edu.au

William Greene
Department of Economics
Stern School of Business, New York University
44 West 4 St., 7-90
New York, NY, 10012
URL: http://people.stern.nyu.edu/wgreene
Email: wgreene at stern.nyu.edu
Ph. +1.212.998.0876
Editor in Chief: Journal of Productivity Analysis
Editor in Chief: Foundations and Trends in Econometrics
Associate Editor: Economics Letters
Associate Editor: Journal of Business and Economic Statistics
Associate Editor: Journal of Choice Modeling

More information about the Limdep mailing list