>
>
> I would like to do an internal validation of a discriminative ability of a
> mixed effects models.
>
> Here is my scrip:
>
> ###########################
> ####bootMer-> boot AUC#####
> ###########################
>
> library(lme4)
> library(lattice)
> data(cbpp)
>
> #fit a model
>
> cbpp$Y<-cbpp$incidence>=1
> glmm<-glmer(Y~period + size + (1|herd), family=binomial, data=cbpp)
> glmm
>
> ##### funcio: versio 3 - no cal posar endpoint en la funcio
> ##########################################################
>
>
>
> AUCFun <- function(fit) {
> library(pROC)
> pred<-predict(fit, type="response")
> AUC<-as.numeric(auc(fit@resp$y, pred))
> }
>
>
> #test
>
> (AUCFun(glmm))
>
> ###run bootMer: AUCFun
>
>
>
> system.time(AUC.boot <- bootMer(glmm,nsim=100,FUN=AUCFun,seed=1982,
> use.u=TRUE,
> type="parametric", parallel="multicore",
> ncpus=2))
>
>
> #...
>
> (boot.ci(AUC.boot, index =c(1,1), type="norm"))
>
> roc(cbpp$Y, predict(glmm, type="response"))
>
>
> #Now it seems more reasonable, bias as "optimism"... but still do not know
> #if I am just doing a AUC with bootstrap CI
> **************************************************************************************************************************************
>
>
>
______________________________________________
R-help@r-project.org mailing list -- To UNSUBSCRIBE and more, see
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
