Have an idea?

Visit Sawtooth Software Feedback to share your ideas on how we can improve our products.

Why does Hierarchical Bayes not allow cross-attribute comparison between attributes?

If I understand correclty, Hierarchical Bayes does not allow cross-attribute comparison between attributes, meaning that one cannot compare the utilities of label levels and the utilities of brand levels and conclude the consumer’s preference.

But why is this the case?
asked May 20, 2019 by Charlotte

1 Answer

0 votes
No standard conjoint method, whether it involves OLS regression, logit, latent class MNL, or HB MNL allows direct comparison of the part-worth (dummy-coded for categorical attribute labels) utility of one level from one attribute to the utility of one level of a different level.  This is due to the experimental design, where one and only one level from each attribute is brought together to form a product concept in the conjoint questionnaire.

You CAN compare the utilities of levels WITHIN the same attribute, allowing you to judge relative preference within an attribute.  

You can also conclude a consumer's preference regarding joint combinations of attribute levels.  For example, you could sum the two part-worth utilities for Red and Convertible (levels from two different attributes) and compare it to the sum of Blue and Sedan to see which sum of utilities is higher (and therefore which product concept the respondent would be more likely to choose).
answered May 20, 2019 by Bryan Orme Platinum Sawtooth Software, Inc. (172,690 points)