Modeling Ordinal Responses for a Better Understanding of Drivers of Customer Satisfaction
Jorge Alejandro, Market Probe, Sharon Kim
While survey researchers make great attempts to standardize their questionnaires – including the usage of ratings scales – oftentimes the methods applied rely on the assumption that the collected response is interval level, as opposed to ordinal. Although analytical approaches that are better equipped to handle ordinal responses have become more accessible over the last several years, many times their use is affected by the requirements of the method, including sample size. Furthermore, often there are sub-populations of interest that are left unexplored because of the additional strain this would impose on the sample size requirement. This paper describes an implementation that uses a Bayesian approach to model an ordinal response while capturing the distribution of heterogeneity in the included sub-populations. More specifically, SAS ® PROC MCMC is used to carry out a comprehensive…