THE DIRICHLET‐MULTINOMIAL MODEL: ACCOUNTING FOR INTER‐TRIAL VARIATION IN REPLICATED RATINGS

摘要: Differences in sensory acuity and hedonic reactions to products lead latent groups pooled ratings data. Manufacturing locations time differences also are sources of rating heterogeneity. Intensity ordered categorical Categorical responses follow a multinomial distribution this can be applied data over trials if the probabilities constant from trial trial. The common test statistic used for comparing vectors proportions or frequencies is Pearson chi-square statistic. When obtained repeated experiments cluster sampling procedure, covariance matrix vector category differ dramatically one assumed model because inter-trial. This effect referred as overdispersion. standard does not fit overdispersed practical implication that an inflated Type I error result seriously erroneous conclusion. Another overdispersion measurable quantity may interest it signal presence segments. Dirichlet-Multinomial (DM) introduced paper intensity Methods estimating parameters DM statistics based on them against specified compare given. A novel theoretical contribution method calculating power tests. useful both evaluating tests determining sample size number trials. goodness extended further Generalized (GDM) model, which multiple variation considered. GDM its applications discussed paper. Applications models consumer research illustrated using numerical examples.