|
Correlation of Estimates Because an individual differences conjoint model was estimated, it is possible to test the correlations of the estimates of the utilities derived from the two methods. These correlations are calculated on the basis of 8 subjects × 16 unique combinations, or 144 distinct cases. The correlations include:
With one exception, attribute 8, all correlations are of the approximate magnitude that we would anticipate. We suspect that part of the problem with the poor across-method correlation for attribute number 8 stems from the lack of precision in the original formulation of the attribute ranking task. As originally formulated, subjects were asked to rank the overall importance of the attribute, not the importance of their choice on the attribute. Using the tea example presented earlier, subjects were asked how important temperature was, not how important their desire for hot tea was. The distinction is a small one, but one we believe is important. |