Bayesian Estimation for Inequality Constrained Analysis of Variance推荐.pdfVIP

3 Bayesian Estimation for Inequality Constrained Analysis of Variance Irene Klugkist and Joris Mulder Department of Methodology and Statistics, Utrecht University, P.O. Box 80140, 3508 TC Utrecht, the Netherlands i.klugkist@uu.nl and j.mulder3@uu.nl 3.1 A Short Introduction to Bayesian Statistics In Chapter 2, several examples of research questions in the analysis of vari- ance (ANOVA) context were presented. The model parameters of an ANOVA are two or more population means and the common and unknown residual variance. In the examples, the hypotheses or research questions of interest impose inequality constraints on the means. For instance, for the four-group ANOVA in the Dissociative Identity Disorder (DID) data from Huntjens et al. [10], the hypothesis µcon µamn {µpat , µsim } represents one of the theories of the researchers. In this chapter, Bayesian estimation of the parameters of inequality constrained ANOVA models is introduced. In the Bayesian approach, knowledge about model parameters is repre- sented by a probability distribution. Two important ingredients of Bayesian analyses are the prior distribution and the posterior distribution of the pa- rameters. The prior represents the knowledge (or uncertainty) about model parameters before observing the data. After observing data, the information in the data is combined with the information in the prior, leading to the pos- terior. Stated otherwise, the posterior distribution represents the knowledge about model parameters after observing the data. In Bayesian analyses, the role of prior distributions is a point of discus- sion. Many prior specifications can be made and different prior distributions may lead to different conclusions. Before moving to the inequality constrained ANOVA model, the basic principles of Bayesian analyses and the role of prior distributions herein wi