logistic regression

[Manuela Gonzalez Suarez]

Hi,

I am trying to run some logistic regressions models that I would like to
compare using AKAIKE criteria, however when I use certain combinations of
variables I get this message

"Warnings:The parameter covariance matrix cannot be computed. Remaining
statistics will be omitted."

followed by

"Model Summary
Step    -2 Log likelihood
1             .000(a)
a    Estimation terminated at iteration number 25 because a perfect fit is
detected. This solution is not unique."

What is happening? I need the "-2 Log likelihood" to calculate AIC and I am
not sure what to do. I don't believe my model is perfect either :-)

Any help will be appreciate it.

Manuela



--
"I have never let my schooling interfere with my education"
                                               Mark Twain
Manuela Gonzalez
SOLS Biology Graduate Program
P.O Box 874601 ASU Tempe, AZ 85287-4601
http://www.public.asu.edu/~lrgerbe/manuela.htm

[Ornelas, Fermin]

Well, there is a problem somewhere in you data or the estimation
process. The variance covariance matrix is based on the matrix of the
predictors weighted by the probability value and if there are problems
to calculate it could be because your matrix is not of full rank i.e.
the matrix could be singular. That is probably why you are getting the
error message that the solution is not unique. If have some predictors
that are indicators or variables that are highly correlated with each
other it could cause the covariance matrix to become singular.

Do some correlation analysis prior to fitting the model and some
descriptive statistics to get a grasp on what your data is doing to your
model. I have built many models in SAS not in SPSS and have never had
any problems with the estimation. Most logistic regressions converge
relatively fast.

Fermin Ornelas, Ph.D.
Management Analyst III, AZ DES
Tel: [phone]
E-mail: [hidden email]

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Manuela Gonzalez Suarez
Sent: Thursday, March 22, 2007 5:59 PM
To: [hidden email]
Subject: logistic regression

Hi,

I am trying to run some logistic regressions models that I would like to
compare using AKAIKE criteria, however when I use certain combinations
of
variables I get this message

"Warnings:The parameter covariance matrix cannot be computed. Remaining
statistics will be omitted."

followed by

"Model Summary
Step    -2 Log likelihood
1             .000(a)
a    Estimation terminated at iteration number 25 because a perfect fit
is
detected. This solution is not unique."

What is happening? I need the "-2 Log likelihood" to calculate AIC and I
am
not sure what to do. I don't believe my model is perfect either :-)

Any help will be appreciate it.

Manuela



--
"I have never let my schooling interfere with my education"
                                               Mark Twain
Manuela Gonzalez
SOLS Biology Graduate Program
P.O Box 874601 ASU Tempe, AZ 85287-4601
http://www.public.asu.edu/~lrgerbe/manuela.htm

NOTICE: This e-mail (and any attachments) may contain PRIVILEGED OR
CONFIDENTIAL information and is intended only for the use of the
specific individual(s) to whom it is addressed.  It may contain
information that is privileged and confidential under state and federal
law.  This information may be used or disclosed only in accordance with
law, and you may be subject to penalties under law for improper use or
further disclosure of the information in this e-mail and its
attachments. If you have received this e-mail in error, please
immediately notify the person named above by reply e-mail, and then
delete the original e-mail.  Thank you.

[Anthony Babinec]

You did not give us information on your sample size or
number of predictors. There are some problematic data
situations for logistic regression illustrated by the
following small example.

Consider the following data

y x z
1 1 1
1 2 2
1 3 3
1 4 4
1 5 5
0 6 5
0 7 6
0 8 6
0 9 7
0 9 8

Consider X as a single predictor of Y. There is a situation
in these data called separability. Note that a cutpoint on
X of 5.5 separates the Ys. Try a logistic regression of
Y on X, and you get Bs and Standard errors that blow up.
Likewise, consider Z as a single predictor. Here, Z=5
straddles the boundary of Y being 0 or 1. A similar thing
happens with Z as a single predictor.

X and Z are highly but not perfectly correlated. When you
do a LR of Y on both X and Z, Logistic Regression prints
the kind of messages that you report.

So, take a closer look at your data, and look for situations
of separability or near-separability and/or high correlation
in your predictors.


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Manuela Gonzalez Suarez
Sent: Thursday, March 22, 2007 7:59 PM
To: [hidden email]
Subject: logistic regression

Hi,

I am trying to run some logistic regressions models that I would like to
compare using AKAIKE criteria, however when I use certain combinations of
variables I get this message

"Warnings:The parameter covariance matrix cannot be computed. Remaining
statistics will be omitted."

followed by

"Model Summary
Step    -2 Log likelihood
1             .000(a)
a    Estimation terminated at iteration number 25 because a perfect fit is
detected. This solution is not unique."

What is happening? I need the "-2 Log likelihood" to calculate AIC and I am
not sure what to do. I don't believe my model is perfect either :-)

Any help will be appreciate it.

Manuela



--
"I have never let my schooling interfere with my education"
                                               Mark Twain
Manuela Gonzalez
SOLS Biology Graduate Program
P.O Box 874601 ASU Tempe, AZ 85287-4601
http://www.public.asu.edu/~lrgerbe/manuela.htm