Full-profile conjoint surveys like CVA are often challenging for respondents to complete. The CVA/HB Module for Hierarchical Bayes Estimation (CVA/HB) module helps you make the most of the precious information you collect. It consistently produces superior results compared to the standard OLS approach.
With CVA/HB, you can reduce the length of your questionnaire and sample sizes without sacrificing precision and accuracy. Precisely the amount of reduction depends on the study and the sample.
How does hierarchical Bayes work? From an intuitive standpoint, HB “borrows” information from the sample population to strengthen estimates of each individual’s part worths. This borrowing of information produces cleaner part worths (fewer reversals, less noise) and actually improves the characterization of unique respondent preferences.
With CVA/HB, you can even field a study in which respondents see fewer conjoint questions than parameters to be estimated. You might ask respondents to complete a random subset of the CVA tasks, using either Web-based data collection or the paper-and-pencil design and data management approach. Academics have shown that you can get useful conjoint information using HB with as few as four conjoint questions per respondent. This “divide and conquer” approach can work well if sample sizes are quite large and individual-level classification of choices is not a priority.
CVA/HB is an integrated component within the SSI Web system.
If you would rather not use SSI Web or SMRT to perform CVA/HB analysis, you may want to review the stand-alone HB-Reg software which also performs OLS (ordinary least squares) regression analysis.