Date: Wed, 15 Dec 2004 09:45:26 -0600
Reply-To: Anthony Babinec <email@example.com>
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: Anthony Babinec <firstname.lastname@example.org>
Subject: Re: Discriminant Analysis
Content-Type: text/plain; charset="us-ascii"
Randomly assign the data to Training(1) or Validation(0) data sets.
There are different ways to do this using COMPUTE, for example.
Then, on the main DISCRIMINANT dialog box, move this created
variable to the Selection Variable entry, and designate 1 as the
Value. DISCRIMINANT will calculate its coefficients based on the
training sample, and will classify both the selected and unselected
cases. Be sure to click the Classify.. button on the main menu
and ask for Summary table.
Having said that, the sample size of 30 is rather small, and your
answer can be contingent on the allocation of cases to Training
and Validation sets. With only 10 cases in the Validation set,
you cannot get a precise estimate of error.
From: SPSSX(r) Discussion [mailto:SPSSX-L@LISTSERV.UGA.EDU] On Behalf Of
Sent: Wednesday, December 15, 2004 4:25 AM
Subject: Discriminant Analysis
Many thanks in advance!
I have run a discriminant analysis based on features(x1, x2 .) for two
classes of an output variable on 30 patterns (data points).
Using pull down menu, flexibility is very limited. I have to use my all
data points for discriminant function. Then leave-one out cross
validation method is used on all same 30 patterns to
gauge the generalization ability of the developed discriminate function
into the classes.
Can I work on this problem in SPSS this way?
Design a model on 20 patterns, measure the performance of function to
classify data items in to classes
After that, use 10 withheld patterns for generalisation of the developed