Thanks in advance for any responses. Please respond to this email
address if at all possible.
I have seen, via Google searches, several old (circa 2006 and earlier)
discussions on the topic: Multinomial Logistic Regression -
Singularities in Hessian Matrix. Honestly, I have not gotten into the
details of the matrix math, but I do recognize that this precludes
calculating standard errors. So, parameter estimates don't converge
and the model build process terminates. So, my questions are:
(1) What is the root-cause of this? Sample size vs. # of vars
(2) How to remedy without totally derailing the research and/or having
to become an SPSS expert or code the entire logic in Matlab?
(3) If SPSS allows the model/analysis to "complete" despite the WARNING
OF SINGULARITIES IN HESSIAN MATRIX, what's the point? That is, if the
underlying mathematics doesn't hold, is there anything salvageable in
the output?
Thank you for any guidance you can offer.
~ Paul
Paul H Ferguson
Lockheed Martin Missiles and Fire Control - Dallas
Systems Engineering
972-603-2919
=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD
Locked
