I have conducted FIMIX and have identified 3 segments.
Post hoc analysis for segment 1 and 2 went fine.
(1) one question here: would it be better to just use the case in the segment with the highest probability / final partition or make an weighted average, i.e. use it in serveral segments with its probability?
for segment 3 I have the following problem: the segment is very "sharp", meaning that some main indicators have always the exact value, i.e. there is no variance.
I now have the problem that FIMIX does not run through.
Can anyone advice what to do?
I have already tried to do some a priori segmentation and try to re-build segment 3's data with a little variance, but the results for this are not really good...
Thx
- Marc