Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: Problem with the cluster-option in relogitq after relogit

From   Mark Schaffer <>
To, joachim Wagner <>
Subject   Re: st: Problem with the cluster-option in relogitq after relogit
Date   Mon, 20 Sep 2004 08:54:56 +0100 (BST)


I think you are right - the estimated probabilities from the no-cluster-
option regression are fine, but the confidence intervals won't be the ones 
you want.

It sounds like a coding problem with relogitq, but let me venture a wild 
guess about the source.  How many clusters (regions) do you have, and how 
many regressors?  For the cluster-robust asymptotics to work, the number 
of clusters has to go off to infinity.  At the other end of the spectrum, 
if the number of regressors is greater than the number of clusters, the 
standard errors you get are not only probably worthless, the var-cov 
matrix on which they're based isn't positive-definite.


Quoting joachim Wagner <>:

> Dear All,
> I am working with the Rare Events Logit program relogit.ado and the
> setx/relogitq - ados (written by Gary King and others). I am using
> an up to 
> date version of Stata 8.2/SE. Friday last week I encountered what
> turned 
> out a rather strange problem for me. I asked Gary King, but he
> replied that 
> "cluster" is a Stata option, not something they in relogit
> programmed, so 
> he can't answer my question. Therefore, I post my question to the
> list:
> I am estimating empirical models to explain the propensity for
> becoming 
> self-employed. I use a large data set for individuals, and ca. 3
> percent of 
> them are about to become self-employed - a rare event. Some of the
> variables in my models are not measured at the level of the
> individuals, 
> but at the level of the regions the people live in. Therefore, I
> estimate 
> the models with the cluster-option to get the estimated standard
> errors 
> right. Everything is going well.
> When I try to use setx/relogitq after estimating a model to
> "forecast" the 
> probability that an individual with a given set of characteristics
> living 
> in a region with a given set of characteristics will become
> self-employed, 
> and when the model is estimated with the cluster-option, I get an 
> error:   r(506) matrix not positive definite.
> Question: Why does this happen, and what can be done?
> As a work-around  I estimate the models without the cluster option;
> given 
> that the estimated coefficients are the same, I think the estimated
> probabilities are fine. But what about the confidence intervals
> reported by 
> relogitq?
> Question: Can I proceed this way?
> I would be very grateful for any comments etc.
> All the best,
> Joachim
> Prof. Dr. Joachim Wagner
> University of Lueneburg
> Institute of Economics
> Campus 4.210
> D-21332 Lueneburg
> Germany
> Phone: +49-4131-78-2330
> Fax: +49-4131-78-2026
> homepage:
> You can access working paper versions of some of my papers on the
> Social 
> Science Research Network (SSRN) at the following URL:
> *
> *   For searches and help try:
> *
> *
> *

Prof. Mark Schaffer
Director, CERT
Department of Economics
School of Management & Languages
Heriot-Watt University, Edinburgh EH14 4AS
tel +44-131-451-3494 / fax +44-131-451-3008


This e-mail and any files transmitted with it are confidential
and intended solely for the use of the individual or entity to
whom it is addressed.  If you are not the intended recipient
you are prohibited from using any of the information contained
in this e-mail.  In such a case, please destroy all copies in
your possession and notify the sender by reply e-mail.  Heriot
Watt University does not accept liability or responsibility
for changes made to this e-mail after it was sent, or for
viruses transmitted through this e-mail.  Opinions, comments,
conclusions and other information in this e-mail that do not
relate to the official business of Heriot Watt University are
not endorsed by it.
*   For searches and help try:

© Copyright 1996–2021 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index