Dear Statalist,
In the very last paragraph Bill mentioned
The logic, "if the number of estimates increases at the same rate as
number of observations, there will be problems" is generally true, the
exception being cases where there is a particular kind of separability,
which happens only in the linear case.
I think another exception is the Poisson case. My reason for saying this
is
1) Hilbe (2007) Page 202
"However, it has been demonstrated by Greene (2006) and others that the
IP problem is not real when applied to the Poisson model. This
conclusion is based on the observation that the Poisson conditional
fixed-effects estimator is numerically equal to the unconditional
estimator, which means that there is no IP problem. On the other hand,
the IP problem does affect the unconditional fixed-effects negative
binomial."
2) Allison (2005) Page 57
"This is not a problem for linear models and (somewhat surprisingly) for
the Poisson models discussed in Chapter 4. But it is a serious problem
with logistic regression and many other nonlinear regression models."
3) Allison (2005) Page 90
"...we find the coefficients for the R&D measures and for the time
dummies are identical for the conditional and unconditional methods.
Standard errors, chi-squares and p-values are also identical for these
variables. Unlike logistic regression, for which conditional and
unconditional estimates can differ substantially, these two methods
always produce identical results for Poisson regression (Cameron and
Trivedi 1998)."
References:
Allison PD (2005) Fixed Effects Regression Methods for Longitudinal Data
Using SAS. SAS Press.
Cameron AC and Trivedi PK (1998) Regression Analysis of Count Data.
Cambridge University Press.
Greene WH (2006) LIMDEP Econometric Modeling Guide, Version 9,
Plainview, NY: Econometric Software Inc.
Hilbe JM (2007) Negative Binomial Regression. Cambridge University
Press.
Best wishes, Garry
Garry Anderson
School of Veterinary Science
University of Melbourne
250 Princes Highway Ph 03 9731 2221
WERRIBEE 3030 Fax 03 9731 2388
Email: [email protected]
-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of William
Gould, StataCorp LP
Sent: Friday, October 26, 2007 12:56 AM
To: [email protected]
Subject: Re: st: conditional logistic
Ricardo Ovaldia <[email protected]> asks,
> What is the difference between conditional logistic regression
> grouping on clinic and unconditional logistic regression including
> clinic as a dummy
> (indicator) variable? That is, what is the difference in model
> assumptions and parameter estimates?
The difference is that the logistic regression estimates are
inconsistent and bad.
--cut--
The logic, "if the number of estimates increases at the same rate
as
number of observations, there will be problems" is generally true,
the exception being cases where there is a particular kind of
separability, which happens only in the linear case.
<end>
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
