Date: Mon, 26 May 2003 07:03:57 -0500
Reply-To: Anthony Babinec <tbabinec@ameritech.net>
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: Anthony Babinec <tbabinec@ameritech.net>
Subject: Re: SPSS 10.x Win DFA interpretation still need HELP!!!!
In-Reply-To: <5.1.1.6.2.20030526053137.00bd6008@pop.starband.net>
Content-Type: text/plain; charset="iso-8859-1"
P(g|d) is the posterior probability of being in group g given the
discriminant score.
For 3 groups, the 3 probs will sum to 1. The case is assigned to the
group with the highest prob.
These are the probabilities you want. The output shows highest group
assignment and
probability, and then the second highest group assignment and
probability. For a 3 group
problem, the third group probability is 1 minus the sum of the first two
group probabilities.
P(d|g) is the likelihood. It is computed under the assumption of
within-groups normality plus some
assumptions about the structure of the covariance matrices.
P(g) is the prior.
Bayes theorem is used to obtain p(g|d) from knowledge of the other two
probs.
Anthony Babinec
-----Original Message-----
From: SPSSX(r) Discussion [mailto:SPSSX-L@LISTSERV.UGA.EDU] On Behalf Of
WCS-Belize
Sent: Monday, May 26, 2003 6:39 AM
To: SPSSX-L@LISTSERV.UGA.EDU
Subject: SPSS 10.x Win DFA interpretation still need HELP!!!!
Importance: High
Hi all,
Running SPSS 10.x Windows. Running DFA and have a question on the
probabilities given in the Case Wise Statistics table. Being way out in
the jungle we have limited resources (no discussion with colleagues) so
all of you out there are the support network :-)
The SPPS results Coach provides an explanation that is 180° from what
was given in a few examples provided on WWW sites using DFA. This is
confusing at best.
So can someone like Arthur J. Kendall help with the explanation of
interpretation of the various probabilities displayed in SPSS when you
run DFA?
The first display is P(D>d| G=g) over p and df, the followed by
P(G=g|D=d) and then the second group P(G=g|D=d) is given.
What is not clear, at least for a dumb field biologist :-) is which of
the probabilities are used to determine which is the predicted group
membership. What is important is to determine when unclassified cases
are classified incorrectly into a class they do not belong to. This is
for cases when they actually do not belong to a predefined or classified
case at all.
Thanks again for all the previous replies that were close but did not
answer the key question.
Cheers from the Jungles of Belize.
Bruce W. Miller
Associate Conservation Zoologist
Wildlife Conservation Society
(founded in 1895 as the New York Zoological Society)
Gallon Jug, Belize
Mailing address
P.O. Box 37, Belize City
Belize, Central America
NEW Phone/Fax +501-220-9002
Visit the
Belize Biodiversity Information System
or the
Neotropical Bat Information System
at
http://fwie.fw.vt.edu/wcs/
OR
Check out the WCS web site at: http://www.wcs.org <http://www.wcs.org/>
OR
Check out information on the Jaguar Conservation Program at
http://www.savethejaguar.com <http://www.savethejaguar.com/>
