Date:         Tue, 18 Nov 2003 09:15:09 -0600
Reply-To:     Anthony Babinec <tbabinec@ameritech.net>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Anthony Babinec <tbabinec@ameritech.net>
Subject:      Re: Discriminant Analysis - Prediction
Comments: To: victor.torres@provencred.com.ar
In-Reply-To:  <000a01c3ade3$c7ceb4e0$ed0ec8c0@provencred.com.ar>
Content-Type: text/plain; charset="us-ascii"

You need to specify either the following

/statistics coeff

or Fisher's coefficients on Discriminant's Statistics check box.

You will get as many sets of coefficients as there are groups
in the target variable.  Turn each of these into a COMPUTE statement.
On a case-by-case base, the maximal score signals the predicted group.

Anthony Babinec

-----Original Message-----
From: SPSSX(r) Discussion [mailto:SPSSX-L@LISTSERV.UGA.EDU] On Behalf Of
Victor Torres
Sent: Tuesday, November 18, 2003 8:54 AM
To: SPSSX-L@LISTSERV.UGA.EDU
Subject: Discriminant Analysis - Prediction


Dear List,
I need to use multinomial discriminant analysis to predict group membership
(I have three groups in the DV) in a new database. I already applied the two
discriminant functions (that I got when I developed the model) to every case
in the new database (so, I have two discriminant punctuations). My doubt is
how to build the rules to predict membership groups from their discriminant
puntuations.

In development, the centroids were:

         function 1    function 2
group 1)   0.144       -8.347E-02
group 2)   0.142        8.386E-02
group 3)  -0.285       -3.939E-04

Any help would be very helpful.

Regards.

Eduardo