Bayesian ranking responses in multiple-response questions

Abstract

Questionnaires are important surveying tools that are used in numerous studies. Analyses of multiple-response questions are not as well established in detail compared with single-response questions. Wang has proposed several methods for ranking responses in multiple-response questions under the frequentist set-up. However, prior information may exist for ranks of responses in numerous situations. Therefore, establishing a methodology that combines updated survey data and past information for ranking responses is an essential issue in questionnaire data analysis. This study develops Bayesian ranking methods based on several Bayesian multiple-testing procedures to rank responses by controlling the posterior expected false discovery rate. Moreover, a simulation is conducted to compare these approaches, and a real data example is presented to show the effectiveness of the methods proposed.