On Tue, 15 Jun 1999, Paul C. Smith wrote:
> Karen -
> > On Tue, 15 Jun 1999, Karen Yanowitz wrote:
> > > However, the depend. variable is yes/no. My understanding of
> > > multiple regression is that it "figures out" the % of variance that
> > > is accounted for by the set of predicotrs- so is it correct to say
> > > that doesn't make sense to use multiple regression - and most
> > > importantly- WHAT WOULD I USE INSTEAD? :)
> >
> > There is a procedure, with which I am not familiar except by
> > name, called logistic regression.
>
> I believe the simplest answer is discriminant analysis, a method described
> by SPSS as follows:
> =========
> This procedure determines the linear combination of predictor variables that
> best classifies cases into one of several known groups. It can use this
> solution to classify cases whose group is unknown.
> =========
It would seem that logistic regression would be appropriate as well. From
SPSS:
"Logistic regression is useful for situations in which you want to be able
to predict the presence or absence of a characteristic or outcome based on
values of a set of predictor variables. It is similar to a linear
regression model but is suited to models where the dependent variable is
dichotomous."
The advantage of logistic regression, though, is that the assumptions
aren't as strict as are those of discriminant analysis (e.g., the
predictors need not be normally distributed or have equal variances in
each group).
Jeff
================================================ ____________________
Jeff Bartel Grad Student in Social Psyc | Manhattan >
[EMAIL PROTECTED] Dept of Psychology | x \_
www-personal.ksu.edu/~jbartel Kansas State U. | |
Syllabi page for psychology instructors: | |
www-personal.ksu.edu/~jbartel/syllabi.html | KANSAS |
================================================ |______________________|
Reading an email message about a new email virus? Getting a note that's
been forwarded to a dozen other people? Before you pass it along, drop by
http://www.US.datafellows.com/news/hoax/ for a list of recent hoaxes and
chain letters.
