ari gold
03-25-2009, 02:03 PM
Hi guys,
I'm having a problem with determing the internal validity of my conjoint model. I have used a fractional factorial design for a full profile ratings based conjoint analysis. 6 variables each with two levels. Respondents needed to rate 10 profiles, from which 2 were holdouts. If I run the conjoint analysis, in SPSS, I get a ,9 for Pearson's R and ,7 for Kendall;s tau. Relative high numbers, indicating a good goodness-of-fit. The kendall's tau value for holdouts reports a 1,000 and a dot at the significance value.
I have read that normally, kendall/s tau for holdouts is lower than the other 2 coeffients. Anybody have an idea what this 1,000 value means, and why it is not lower than the other two.
Second question, it is mentioned by different authors/market research books that correlation coefficients between the predicted values of these holdout observations and their actual values can be used to assess the internal reliability/validity of the model.
But how do i do this; do I take the calculated utilities according to the holdout profiles and correlate this with the actual ratings.
Any insight is highly appreciated
I'm having a problem with determing the internal validity of my conjoint model. I have used a fractional factorial design for a full profile ratings based conjoint analysis. 6 variables each with two levels. Respondents needed to rate 10 profiles, from which 2 were holdouts. If I run the conjoint analysis, in SPSS, I get a ,9 for Pearson's R and ,7 for Kendall;s tau. Relative high numbers, indicating a good goodness-of-fit. The kendall's tau value for holdouts reports a 1,000 and a dot at the significance value.
I have read that normally, kendall/s tau for holdouts is lower than the other 2 coeffients. Anybody have an idea what this 1,000 value means, and why it is not lower than the other two.
Second question, it is mentioned by different authors/market research books that correlation coefficients between the predicted values of these holdout observations and their actual values can be used to assess the internal reliability/validity of the model.
But how do i do this; do I take the calculated utilities according to the holdout profiles and correlate this with the actual ratings.
Any insight is highly appreciated