Hierarchical Bayes Estimation in Conjoint Analysis
The use of Hierarchical Bayes (HB) estimation in conjoint analysis is all the craze in the marketing world. Why is this? What is Hierarchical Bayes estimation? How does a different kind of analysis technique improve the results of a conjoint analysis study?
Hierarchical Bayesian estimation is a complex but powerful approach of modeling data sets to yield more precise and granular analysis. The HB model internalizes prior probabilities and data-produced likelihoods to compute posterior probabilities it an iterative process. Its methods are optimized by utilizing Markov Chain Monte Carlo (MCMC) simulations as a means of estimation.
HB estimation is being used in conjoint analysis to determine the partworth utilities because it does so more accurately than other linear or logit models. This along with the fact that HB estimation has the recovery ability to calculate these more precise results while showing fewer packages has created the excitement about HB estimation in conjoint analysis.
In the context of conjoint analysis, HB estimation takes into account the prior knowledge of the features, the individual’s preference selections as well as the preferences of all who participated in the survey to derive preference scores. The mathematics driving the MCMC simulation allows the process to borrow information from the full data set to estimate the partworth utilities, providing estimates in situations where classical methods fall short.
There are many benefits and opportunities that are introduced into conjoint analysis with the use of HB estimation. It opens avenues to all models of conjoint analysis, especially variations of choice-based conjoint analysis (CBC) and adaptive conjoint analysis (ACA). No longer do marketers have to bombard respondents to question after question of packages to examine and rank. This leads to respondent fatigue and skewed results. With HB estimation, many statistical conditions of commonly used regression models are no longer required and allow the researcher to get amazing, more truthful results without the risk of boring the survey-taker with long questionnaires.
Qualtrics conjoint team has researchers with backgrounds in mathematics and statistics who spent years mastering HB estimation with conjoint analysis. From start to finish, they understand the entire HB process and have the knowledge to leverage this powerful analysis method for conjoint analysis. They have been using HB estimation to derive preference scores for conjoint analysis for years.