J. A. Cano Sánchez, C. Carazo Díaz, D. Salmerón Martínez
An objective Bayesian model selection procedure for one way analysis of variance under heteroscedasticity was proposed in Bertolino et al. (2000) and, and, recently, specific solutions for the special case when homoscedasticity is present have been developed in Cano et al. (2011). In this paper a solution based on an objective Bayesian model selection procedure is proposed for linear contrasts between means for the one way ANOVA in the context of both homoscedasticity and heteroscedasticity. Bayes factors for intrinsic priors are used instead of Bayes factors for the usual default prior distributions since for them Bayes factors are not well defined. The classical p-value and posterior probability of the null hypothesis are compared through so-called calibration curves. The behavior of these calibration curves as a function of the sample size is studied too and some interesting conclusions are drawn.
Palabras clave: linear contrasts between means, model selection, Bayes factor, intrinsic priors, robustness, calibration curve
Programado
XC1a Pósters (Estadística)
18 de abril de 2012 12:00
Salón Madrid
