PROBLEM: PROC LOGISTIC (aka PROC LOGIST) terminates with an error
message as follows:
ERROR: Floating Point Overflow.
NOTE: The SAS System stopped processing this step because of errors.
CAUSE: In most cases, the floating point overflow message in
PROC LOGISTIC is a result of the maximum likelihood solution
not existing. This will cause some of the parameters to
become infinite and the procedure fails to converge. It is
common for the maximum likelihood solution to not exist
when there are few or no repeated observations at each
observed setting of the covariates.
SOLUTION: The solution to this problem is to either simplify
the model or make the data more dense by categorizing covariates.
(NOTE: The explanation and solution shown above are repeated verbatim
from eMail received from a SAS Institute statistician on
6 October 1998.)