Date:         Tue, 30 Oct 2007 23:33:56 -0000
Reply-To:     matthewpharr@GMAIL.COM
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Matthew <matthewpharr@GMAIL.COM>
Organization: http://groups.google.com
Subject:      Convert Discriminant Score to Probability
Comments: To: sas-l@uga.edu
Content-Type: text/plain; charset="iso-8859-1"

List users,
I have run a discriminant analysis in SPSS and would like to use the
discrim
score to transform the prior probabilities to posterior probabilities.
I've
run a discrim against a 7 level target. However, I need to score a
separate
dataset with the equation I created. I would like to not only have the
score
in my different dataset but also the probability that SPSS produces. I
have no
problem computing a score with the discrim equation using Fisher's
classification function. If I score a new record that hits my database
with the 7 level discrim equation, I get seven scores
for that record. I would also like to get 7 new variables that track
the
probability of being in each group (summing all probabilities would
equal
1). Has anyone encountered a task like this? I know SPSS uses Bayes'
theorem
to do this, but I can't quite get the formula to work on my new
records.
Also, I have to transport this solution into another application. I
can't simply
bring my new data back into SPSS and run the discrim again. Thanks in
advance for any help provided.

Best,
Matt