Date:         Thu, 5 Jun 2008 16:56:49 -0400
Reply-To:     "Whanger, J. Mr. CTR" <James.Whanger@med.navy.mil>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         "Whanger, J. Mr. CTR" <James.Whanger@med.navy.mil>
Subject:      Re: logistic regression
Comments: To: azam.khan@utoronto.ca
In-Reply-To:  A<20080605163044.d1xg5li74kwk4sw0@webmail.utoronto.ca>
Content-Type: text/plain; charset="us-ascii"

Azam,

There are two additional issues to consider: (1) The proportional
distribution of cases in each of the binary categories.  If the
distribution of the binary variable is not reasonably close to 50/50,
then logistic is more powerful. (2) The distribution of the predictor
variables.  If they are not reasonably close to normally distributed,
then again logistic is the more powerful method.  However, if the binary
variable is close to a 50/50 distribution and the predictors are
reasonably close to normally distributed, then they will provide quite
similar results.

Regards,

Jim

-----Original Message-----
From: SPSSX(r) Discussion [mailto:SPSSX-L@LISTSERV.UGA.EDU] On Behalf Of
azam.khan@utoronto.ca
Sent: Thursday, June 05, 2008 4:31 PM
To: SPSSX-L@LISTSERV.UGA.EDU
Subject: Re: logistic regression

Thanks Paul! Just wondering how many cases would you consider lots,
because some of these data sets I receive are pretty big. And in the
case where I do haev enough data to use discrimant analysis, would you
then recommend I use that instead of logistic? Thanks!


Regards,

Azam


Quoting "Swank, Paul R" <Paul.R.Swank@uth.tmc.edu>:

> If you're going to use discriminant analysis then you need a lot of
> data. Logistic would be the best plan.
>
> Paul R. Swank, Ph.D.
> Professor and Director of Research
> Children's Learning Institute
> University of Texas Health Science Center - Houston
>
>
> -----Original Message-----
> From: SPSSX(r) Discussion [mailto:SPSSX-L@LISTSERV.UGA.EDU] On Behalf
> Of jimjohn
> Sent: Thursday, June 05, 2008 3:07 PM
> To: SPSSX-L@LISTSERV.UGA.EDU
> Subject: logistic regression
>
> If I have a dichotomous categorical dependent variable (ie the only
> values it can take are 1=yes and 0=no), and my independent variables
> are all continuous (not categorical), then is logistic regression the
> best method to use? or should i be using discriminant analysis in this

> case? thanks!
> --
> View this message in context:
> http://www.nabble.com/logistic-regression-tp17678490p17678490.html
> Sent from the SPSSX Discussion mailing list archive at Nabble.com.
>
> =====================
> To manage your subscription to SPSSX-L, send a message to
> LISTSERV@LISTSERV.UGA.EDU (not to SPSSX-L), with no body text except
> the command. To leave the list, send the command SIGNOFF SPSSX-L For a

> list of commands to manage subscriptions, send the command INFO
> REFCARD
>

=====================
To manage your subscription to SPSSX-L, send a message to
LISTSERV@LISTSERV.UGA.EDU (not to SPSSX-L), with no body text except the
command. To leave the list, send the command SIGNOFF SPSSX-L For a list
of commands to manage subscriptions, send the command INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
LISTSERV@LISTSERV.UGA.EDU (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD