[Limdep Nlogit List] (no subject)

wgreene at stern.nyu.edu wgreene at stern.nyu.edu
Sat Feb 21 00:29:29 EST 2009


Craig. TRUNCATION does not support that model. But you can fit it by simulation. Add
;rpm;fcn=one(n);pts=100;halton
To the command.
/Bill Greene

Sent from my Verizon Wireless BlackBerry

-----Original Message-----
From: Craig Gallet <cgallet at sbcglobal.net>

Date: Thu, 19 Feb 2009 12:30:57 
To: Limdep and Nlogit Mailing List<limdep at limdep.itls.usyd.edu.au>
Subject: Re: [Limdep Nlogit List] (no subject)


Thank you for your reply, Professor Greene.  Even with this correction, I still get the same results.  The command structure is:
 
regr;lhs=one;rhs=one;panel;str=scode$
truncation; LHS=Y; RHS=ONE, X, W, Z;
pds=_GROUPTI; random effects; limit=0 $
 
If 200 is too many periods, is it possible that I won't get an error message (i.e., it will simply stop and report the non-panel results)?  
 
Craig


--- On Thu, 2/19/09, William Greene <wgreene at stern.nyu.edu> wrote:

From: William Greene <wgreene at stern.nyu.edu>
Subject: Re: [Limdep Nlogit List] (no subject)
To: "Limdep and Nlogit Mailing List" <limdep at limdep.itls.usyd.edu.au>
Date: Thursday, February 19, 2009, 11:31 AM

Craig.
All models save the linear regression require ;PDS=... not
STR=... You can get the PDS variable as follows:
REGR;LHS=ONE;RHS=ONE;PANEL;STR=SCODE$
after you do this, the variable_GROUPTI is what you need.
Replace ;STR=SCODE with ;PDS=_GROUPTI.
However, you will have a different problem. The estimator will
probably tell  you that 200 is too many periods. The log likelihood
must be computed as the product of the densities, and this is likely
to cause overflow or underflow in your case.
/Bill Greene

----- Original Message -----
From: "Craig Gallet" <cgallet at sbcglobal.net>
To: limdep at limdep.itls.usyd.edu.au
Sent: Thursday, February 19, 2009 1:24:31 PM GMT -05:00 US/Canada Eastern
Subject: [Limdep Nlogit List] (no subject)

Hello,
 
I am trying to estimate a random effects truncated regression model with an
unbalanced panel (the group sizes range from 1 to 200, with over 400 groups) but
am unable to do so.  Specifically, I use commands akin to:
 
Truncation; Lhs = y; Rhs = one, x, z, w,...; str=scode; random effects; limit=0
$
 
Unfortunately, the only output I get are the non-panel truncated results, with
no error messages.  Most of the independent variables are dummies.  Is there a
reason for Limdep to not provide the random effects results, but also to not
provide an error message (e.g., failure to converge)?  Any assistance is
greatly appreciated.
 
Craig       
_______________________________________________
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
_______________________________________________
Limdep site list
Limdep at limdep.itls.usyd.edu.au
http://limdep.itls.usyd.edu.au


More information about the Limdep mailing list