Returns a table of model comparison statistics
compare( ... , nobs=NULL , sort="AICc" , BIC=FALSE , DIC=FALSE , delta=TRUE ,
DICsamples=1e4 )A series of fit models, separated by commas
Number of observations to use in calculating AICc/BIC
Sort table by ascending values in named column
Compute BIC as well
Compute DIC as well. See details.
Include differences between AICc/BIC/DIC values
Number of posterior samples to use in computing DIC
This function computes AICc and optionally BIC and DIC values for fit models and returns a table sorted by ascending values. Each row in this table is a model, and the various columns provide AICc, BIC, DIC, numbers of parameters, and model weights.
At the moment, DIC only works for models fit with map. DIC is computing using the original Dhat + 2(Dbar-Dhat) formulation, not the variance formulation. Samples are drawn from the posterior, assuming it is multivariate gaussian and defined by the variance-covariance matrix of the MAP estimates. The column pD in the results is the effective number of parameters, Dbar-Dhat.
McElreath 2011, Statistical Rethinking.