Statalist The Stata Listserver


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: Fwd: Blogit vs logit and cc


From   Joseph Coveney <jcoveney@bigplanet.com>
To   Statalist <statalist@hsphsun2.harvard.edu>
Subject   Re: st: Fwd: Blogit vs logit and cc
Date   Mon, 05 Jun 2006 21:42:53 +0900

Dominic Muston wrote:

I feel embarassed to ask this question, but here goes. I'm trying to
understand why -blogit- doesn't produce the odds ratio I would expect,
but -logit- (with a weighting option) and -cc- do.

I get the sort of simple dataset I'm exploring by doing this:

use http://www.stata-press.com/data/r9/xmpl2, clear
gen n0=pop-deaths
rename deaths n1
collapse (sum) n0 (sum) n1, by(exposed)
reshape long n, i(exposed) j(outcome)
list

exposed outcome n
0 0 184
0 1 21
1 0 174
1 1 30

where "exposed" is an exposure indicator, "outcome" is an outcome
indicator and "n" is the number of patients.

The command "cc outcome exposed [fweight=n]" and "logit outcome
exposed [weight=n], or" produce an odds ratio of 1.51 with a chi^2_1
value of 1.87 (ie z=1.36). This is entirely what I would expect.

But "blogit outcome n exposed, noconstant or" produces an odds ratio
of 0.005 (z=5.30) and "blogit exposed n outcome, noconstant or"
produces an odds ratio of 0.02 (z=3.87), neither of which matches the
"cc" or "weighted logit" result.

I would expect both -blogit- calls to produce the results that I see
from the -cc- and -logit- calls, but I don't. Why not? (It's going to
be something really obvious isn't it? I did search statalist before
posting.)

--------------------------------------------------------------------------------

-blogit- doesn't really use frequency weights in the same sense of other Stata commands'.





© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index