Hi Alice:
Say we have two explanatory variables, x1 and x2, and these variables
are highly positively correlated. So whenever we see a high value of
x1 we also see a high value of x2. Now how can we say that a change
in y is caused by either x1 or x2? We obviously can't distinguish
between the effect of x1 and x2 if the correlation between x1 and x2
is perfect. If the correlation is close to 1 (or -1) it is extremely
difficult, and the confidence intervals will reflect that by being
very large. The estimates will also tend to be unstable, i.e. small
changes in your data or model will lead to large changes in the
estimates. This is true for any type of multiple regression,
including OLS and Poisson. So using poisson regression will not solve
your multicolinearity problem.
Hope this helps,
Maarten
-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-
statalist@hsphsun2.harvard.edu]On Behalf Of ALICE DOBSON
Sent: maandag 26 september 2005 11:45
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: Multicollinearity in a simple Poisson model
Because multicollinearity may influence the least squares estimates.
This is one of the basic assumptions in OLS. However, I have not come
across such an assumption for Poisson regression [which assumes a
Poisson distribution]. This may be due to my limited exposure to
statistics literature.
*
* 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/