Technical Papers  Pricing & Ordering  Your Cart

HB Estimation

A typical challenge we face as researchers is to estimate a variety of weights (such as utility scores, coefficients, or attribute importances) using a limited amount of data. A relatively new statistical methodology called hierarchical Bayes (HB) improves these estimates, leading to greater stability and validity. HB is commonly used to improve conjoint analysis utilities (for all major conjoint techniques), and to permit individual-level estimation from sparse CBC (Choice-Based Conjoint) data. It may also be applied to MaxDiff scaling, or to general regression-based problems (where respondents have provided multiple cases or observations). Although the mathematics behind HB are very complex, our software makes it easy for researchers to obtain excellent results using robust default settings.

Sawtooth Software offers the following HB software tools:

CBC/HB (for CBC and other user-supplied discrete choice problems)
ACA/HB (for Adaptive Conjoint Analysis)
CVA/HB (for traditional full-profile conjoint)
HB-Reg (for general regression-based problems)

You don't need to have used our tools for designing the questionnaire and collecting the data. Our CBC/HB and HB-Reg tools may be used for data you have collected under third-party applications and procedures.

For introductory articles on HB, please see:

• Hierarchical Bayes: Why All the Attention?
• Understanding HB: An Intuitive Approach

© 2012 Sawtooth Software, Inc. All rights reserved.