Date: Wed, 16 Jul 1997 23:38:59 GMT
Reply-To: Ted Morrison <echess@IX.NETCOM.COM>
Sender: "SPSSX(r) Discussion" <SPSSX-L@UGA.CC.UGA.EDU>
From: Ted Morrison <echess@IX.NETCOM.COM>
Organization: U.A.B. School of Public Health
Subject: Re: strength of association in logistic regression
If you take initial -2 log likelihood and subtract the -2 log
likelihood, then divide this difference by the initial log likelihood
you will get the percent improvement in the fit of the logistic curve.
I have no reference for this. This was told to me by a political
sciece professor when I was writing my master's thesis in public
health. If anyone has any comments on this procedure please enlighten
1:03 -0400, "Popovic, Jennifer, VBAVACO" <ORMJPOPO@VBA.VA.GOV> wrote:
>To: Netnews Server <NETNEWS@AMERICAN.EDU>
>I need to calculate a statistic using logistic regression that is
>analogous to the r-square statistic in linear regression.
>Any suggestions on how to go about doing this?
>Surveys and Research Staff
>Veterans Benefits Administration
>firstname.lastname@example.org (Internet email)