The PHREG Procedure |
Denote the observed data by . Let be the vector of parameters of length . Let be the likelihood. The deviance information criterion (DIC) proposed in Spiegelhalter et al. (2002) is a Bayesian model assessment tool. Let Dev. Let and be the corresponding posterior means of and , respectively. The deviance information criterion is computed as
Also computed is
where is interpreted as the effective number of parameters.
Note that defined here does not have the standardizing term as in the section Deviance Information Criterion (DIC). Nevertheless, DIC calculated here is still useful for variable selection.
Copyright © 2009 by SAS Institute Inc., Cary, NC, USA. All rights reserved.