Dear all,
I 'm doing a conjoint analysis in spss and I
have the following problem:
I know in advance the relationships in the
levels of the factors (This is
confirmed by expert people who are also doing
the ranking of cards and any
other relationship maybe anacceptable by them).
If I use the (disctrete) model for the levels I get some relationships that
are
not in the right direction but the Pearson' R is really great both
for
design cases and holdouts (0.995 and 0.667 respectively). If I postulate
a
linear model the problem is fixed but Pearson's R goes down (0.65 and
0.33
respectively).
One option is to use the latter model but I
wonder if this problem indicates
that something is wrong with the ranking of
cards or the number of cards in
the orthogonal is not adequate (27 design + 4
holdouts=31 cards for 10
factors with 3 levels each).
Is there any
possible solution?
I 'm also interested in a good book in Conjoint
Analysis.
Thanks in
advance
Christos