Dear all--

I have data from a Conjoint questionnaire -- 3 attributes with 3 levels each; design generated by Orthoplan: 9 cards + 1 holdout; spl size N=600 -- and get the following results:

Kendall's tau = 1,000 Significance = ,0001 Kendall's tau = ,000 for 1 holdouts Significance = ,

I learned that the correlation between estimates for holdouts and their 'real' ranking is interpreted as model fit. According to this, my model must be bad -- Tau=0. Hard to believe, because Tau for all other cards is very high.

Does maybe anyone have a reference about how Tau is computed in SPSS? Or does it anyway need more than 1 holdout to compute Tau for holdouts?

Thank you very much, greetings,

--Peter

----------------------------------------------------------------------- P.S Lubisz oglądać reklamy? Obejrzyj i zagłosuj na najlepszą! < http://tytany.wp.pl/wt.html >