A quick addition:
SAS PROC LOGISTIC with option SCALE=PEARSON only calculates what R
calculates in the grouped version; this is forced by refusing to
calculate a FI estimate for ungrouped data, unless an AGGREGATE option
is specified.
Apparently, SAS has decided for one model (not the one I
Dear list,
in preparing a lecture, I have created an example for the same data in
grouped and ungrouped form (the well-known Titanic data). The code
included below shows that the overdispersion estimates are strongly
different for the two approaches.
If the overdispersion parameter FI is seen
