[Limdep Nlogit List] wheighting and choice based sampling

William Greene wgreene at stern.nyu.edu
Tue Mar 10 04:13:19 EST 2009

That is not going to be possible.
"Weighting" in nlogit is handled at the individual level, not the
choice level. It's possible to trick the program into weighting
the individual choices by switching to "frequency" data where the
frequencies are the weights. But, if you do that, you will hide the
ranks - it will not be possible to discern what the ranks of the
observations are. Rank data need to be based on individual data,
so it is not possible to provide choice specific weights.
/B. Greene

----- Original Message -----
From: scolombo at ugr.es
To: limdep at limdep.itls.usyd.edu.au
Sent: Monday, March 9, 2009 12:11:58 PM GMT -05:00 US/Canada Eastern
Subject: [Limdep Nlogit List] wheighting and choice based sampling

Hi to all,

I need to estimate a conditional logit model (with ranked choices), where
I need to weight each individual choice with different weights. In the
limdep manual it is advised to use a weight variable (;wts=name) (ref.=
10-12 reference guide Nlogit 4.0).

I created a weight variable that contains the weights for each individual
and observation, but the model seems to ignore it. (results are the same
than the ones without weights).

I wonder if somebody had the same problem/need in the past and can advise
on that.

Thanks in advance.


Limdep site list
Limdep at limdep.itls.usyd.edu.au

More information about the Limdep mailing list