[Limdep Nlogit List] re: panel tobit model

[William Greene]

Ms. Qui.  I cannot comment on the correctness of your model
for your data.  Perhaps an ordered probit model might be better
since it looks like the outcome is an ordered, but qualitative
measure.  As for the RE model, if the data give no indication of
correlation of the unobservables, then a cross section approach
might be better.  On the other hand, it seems to me that given what
you are measuring, for most observations, there will be no variation
across time in the LHS variable.  That makes it very difficult to fit
any kind of panel data model.
/B. Greene

************************************************
Professor William Greene
Department of Economics
Stern School of Business
New York University
44 West 4th St., Rm. 7-78
New York, NY   10012
Ph. 212.998.0876
Fax. 212.995.4218
URL. http://www.stern.nyu.edu/~wgreene
Email. wgreene at stern.nyu.edu
************************************************

----- Original Message -----
From: ecptq at bath.ac.uk
Date: Sunday, May 20, 2007 6:10 pm
Subject: [Limdep Nlogit List] re: panel tobit model

> Dear Professor Greene,
> 
> Thanks a lot for your quick reply. I double checked the grpsize 
> variable, it is correct, although this is an unbalance panel.
> 
> The value range of the left hand side variable is from 0 to 12, 
> and the 
> value '0' indicates a good body shape. As to other values, the 
> more 
> they are away from'0', the worse is the body shape.
> 
> So I fit the (random-effects) tobit model with the default 
> censoring 
> specification which is left censoring at 0.0, as the estimation of 
> a 
> fixed effects Tobit model is problematic as argued in your paper.
> 
> Based on this, can you kindly suggest me (i) whether the tobit is 
> the 
> right model to select; (ii), if so, the random-effects is the only 
> option, isn't it?
> 
> Thanks a lot.
> 
> Best regards
> Tian Qiu
> 
> Dear Ms. Qui:  Assuming the grpsize variable is set up correctly, then
> (1) No, with only two iterations, the RE model has not been 
> estimated and
> (2) You should rethink the random effects specification.  This 
> outcome is
> not due to a failure of the program. It occurs if the RE model is 
> the wrong
> model for the data.
> Regards,
> Bill Greene
> 
> ************************************************
> Professor William Greene
> Department of Economics
> Stern School of Business
> New York University
> 44 West 4th St., Rm. 7-78
> New York, NY   10012
> Ph. 212.998.0876
> Fax. 212.995.4218
> URL. http://www.stern.nyu.edu/~wgreene
> Email. wgreene at stern.nyu.edu
> ************************************************
> 
> ----- Original Message -----
> From: ecptq at bath.ac.uk
> Date: Sunday, May 20, 2007 3:24 pm
> Subject: [Limdep Nlogit List] panel tobit model
> 
> 
> [Hide Quoted Text]
> > Hi, everyone,
> >
> > I am running the random-effects tobit as follows:
> >
> > tobit;lhs=body;rhs=one,X;pds=grpsize$
> >
> > It comes the results with:
> > Error:   806: Line search does not improve fn. Exit iterations. 
> Status=3> and the table shows iterations completed = 2.
> >
> > Are the results of this random-effects tobit still reliable, and is
> > there anything I can do about it?
> >
> > Thanks in advance.
> 
> -- 
> Miss Tian Qiu
> Department of Economics and International Development
> University of Bath
> Tel.: (01225) 384864
> 
> _______________________________________________
> Limdep site list
> Limdep at limdep.itls.usyd.edu.au
> http://limdep.itls.usyd.edu.au
>