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

Re: st: Lrtest versus Test

From   Richard Williams <>
To, Statalist <>
Subject   Re: st: Lrtest versus Test
Date   Mon, 22 Dec 2003 00:08:02 -0500

With logistic regression, especially, I believe that the likelihood ratio test
is generally preferred: there is an uncommon set of circumstances with
logistic regression that lead to a likelihood function that renders the Wald
test wildly inaccurate. It's known as the Hauck-Donner Phenomenon or Hauck-
Thanks Joseph. I won't claim to fully understand this after a quick read, but statements like "a small t-value could indicate "either an insignificant OR a very significant effect" certainly give reason for concern! If I'm reading this correctly, the truly paranoid should not rely on the z statistics reported by the logistic routine. Instead, you should estimate a series of constrained models where you drop one var at a time and compare to the model with all the vars in. Incidentally, SPSS's Nomreg routine with the LRT option will do this for you, so that is one area where the competition may have a leg up on Stata. (This is computationally intensive, however, and I think SPSS says somewhere that it could be a good time to go get coffee if you are going to use this option.)

So, it would be nice if some future edition of Stata

(a) would give you the option of reporting the LR statistics for each coefficient, just like SPSS will let you do now. (I imagine this would be a fairly simple ado file? Has somebody maybe done it already?)

(b) had an LrTest command that didn't require you to first estimate the constrained model yourself; it would just do it for you.

Richard Williams, Associate Professor
OFFICE: (574)631-6668, (574)631-6463
FAX: (574)288-4373
HOME: (574)289-5227
EMAIL: Richard.A.Williams.5@ND.Edu
WWW (personal):
WWW (department):

* For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index