[Limdep Nlogit List] multivariate probit and selection

elsa marques elsa.r.marques at gmail.com
Fri Oct 16 22:17:20 EST 2009


Dear Wiji,

After they say "yes", can they still choose to give to zero charities? If
so, a zero-inflated model could be possible (ie ZI neg bin/poisson). I used
this type of model a few months back in Limdep. I was using panel data but I
couldn't get estimates for the zero-inflated part of the regression, but it
works very well if you're doing it cross-section and you get the same as
stata. If you end up using this in a panel setting, please let me know if
you find the right comand to retrieve the ZI parameters. If they can't
choose to give to 0 charaties after saying yes, then this model would not be
appropriate.

Elsa

On Fri, Oct 16, 2009 at 10:53 AM, Arulampalam, Wiji <
Wiji.Arulampalam at warwick.ac.uk> wrote:

> Dear All,
>
> Does anyone know of any published references to the following problem
> please?  I would also like to know whether Limdep/Nlogit would able to
> handle the estimation.  The problem is explained below with an example.
>
> In the first stage, individuals decide on whether to contribute to a
> charity or not.  In the second stage, given yes, the individuals choose
> which charity to give.  They can give to more than one charity.  The
> problem I have is that the second stage is only defined for all those
> who chose to go down the route of a 'yes'.  This means that if I used a
> multivariate probit for the second stage, what do I do with the
> probability attached to the outcome of zeros in all of the three
> probits?
>
> I hope you all can understand what I am trying to say here.
>
> Thanks very much.
> Best
> wiji
>
> _______________________________________________
> Limdep site list
> Limdep at limdep.itls.usyd.edu.au
> http://limdep.itls.usyd.edu.au
>


More information about the Limdep mailing list