Research On The Hierarchical Bayes Models In The Application Of Consumer Behavior

In the field of statistics, Bayesian statistic is emerging more and more academic recognition andattention, especially for the Hierarchical Bayesian modeling approach which is one of the mostfamous Bayesian statistical methods in social sciences. For the lack of Hierarchical Bayesian methodsstudy in current local academic and the problems in the consumer behavior research, this paper armsto build a Hierarchical Bayesian random effects model to achieve more accurate estimates of theindividual-level consumer behavior and to describe the differences between consumer preferenceswhich the traditional conjoint models could not reach before.In this paper, first of all, it describes the theory, the basic principles, the inference methods andthe controversy of the Bayesian methods, intended to elaborate the basic idea of the Bayesian theorywithin Hierarchical Bayesian modeling approach; Secondly, this paper describes the modelingmechanism, the computational methods and the various applications fields of the HierarchicalBayesian modeling approach; Thirdly, combined with the deficiencies and difficulties for traditionalconjoint analysis to model the consumer behavior, this paper proposes to build a HierarchicalBayesian random effects model using hierarchical Bayesian modeling approach; Finally, by using ofempirical research method and evaluation research method, this paper conducts the empirical analysisfor Hierarchical Bayesian random effects model in the consumer behavior analysis. By using ofmobile phone products conjoint program which contains six important properties,17properties ofcontour levels and18kinds of products, we get the189sample mobile phone consumers withdifferent demographic variables. Ultimately within the comparisons with traditional dummyvariable regression model, the paper gets the completely analysis on the consumer’s within-unitbehavior and across-unit behavior.According to the empirical results, we get the following conclusions: on the one hand, thehierarchical Bayesian model have much better performance on the fit and the forecasting ability onthe individual-level estimation than traditional dummy variable regression model, and thedemographic variables improve both two abilities at the same time; On the other hand, theHierarchical Bayesian model can describe the differences in the consumers behavior using acontinuous normal distribution and the variable demographic characteristics can explain thesedifferences, so it can explain all the uncertainty that included in the behavior of the consumerpreference.