## [16c8b4]: man / diffdic.Rd  Maximize  Restore  History

### 51 lines (48 with data), 1.6 kB

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 \name{diffdic} \alias{diffdic} \title{Differences in penalized deviance} \description{ Compare two models by the difference of two \code{dic} objects. } \usage{ \special{dic1 - dic2} diffdic(dic1, dic2) } \arguments{ \item{dic1, dic2}{Objects inheriting from class dic''} } \details{ A \code{diffdic} object represents the difference in penalized deviance between two models. A negative value indicatest that \code{dic1} is preferred and vice versa. } \value{ An object of class diffdic''. This is a numeric vector with an element for each observed stochastic node in the model. The \code{diffdic} class has its own print method, which will display the sum of the differences, and its sample standard deviation. } \note{ The problem of determining what is a noteworthy difference in DIC (or other penalized deviance) between two models is currently unsolved. Following the reults of Ripley (1996) on the Akaike Information Criterion, Plummer (2008) argues that there is no absolute scale for comparison of two penalized deviance statistics, and proposes that the difference should be calibrated with respect to the sample standard deviation of the individual contributions from each observed stochastic node. } \author{Martyn Plummer} \references{ Ripley, B. (1996) \emph{Statistical Pattern Recognition and Neural Networks.} Cambridge University Press. Plummer, M. (2008) Penalized loss functions for Bayesian model comparison. \emph{Biostatistics} doi: 10.1093/biostatistics/kxm049 } \seealso{\code{\link{dic}}} \keyword{models}