I have a sample divided into two segments: those who are willing to buy a product, and those who are not willing to buy it.
On the other hand, I have a MaxDiff with a list of k product attributes. My intention is to do a discriminant analysis with the MaxDiff scores (ZCRS) to see how much each of those attributes discriminates between the two purchase intent groups.
My problem is that, since for each respondent the ZCRS always add up to zero, I only have k-1 linearly independent attributes. Therefore, when running the discriminant analysis, one of the k attributes is always left out, since it does not pass the tolerance test.
Is there a way I can get around that problem, and run the discriminant analysis in such a way that it considers all the k attributes of the MaxDiff?