customUniCox: Univarate shrinkage prediction in the Cox model
In bernau/survHDExtra: Addtitional Features for package survHD

Description Usage Arguments Value Author(s) References See Also Examples

Usually, this function is not called directly but embedded in a call to customSurv from the main package survHD (see examples).

See http://cran.r-project.org/web/packages/uniCox/index.html for more details

1	customUniCox(Xlearn, Ylearn, learnind,...)

`Xlearn`	Gene expression data (a `data.frame`)
`Ylearn`	Survival Response, an object of class `Surv`.
`learnind`	An index vector specifying the observations that belong to the learning set. May be `missing`; in that case, the learning set consists of all observations and predictions are made on the learning set.
`...`	Further arguments that shall be passed to `uniCox` function from the uniCox R package (http://cran.r-project.org/web/packages/uniCox/index.html)

An object of class ModelLearned.

Levi Waldron lwaldron@hsph.harvard.edu Christoph Bernau bernau@ibe.med.uni-muenchen.de

Tibshirani RJ: Univariate shrinkage in the cox model for high dimensional data. Stat Appl Genet Mol Biol 2009, 8:Article21.

learnSurvival

##load example data
if(require(pensim)){
library(survHD)
library(survHDExtra)
data(beer.exprs,package="pensim")
data(beer.survival,package="pensim")
##create Surv object (y)
beerY <- with(beer.survival,Surv(os,status))
##create mini X matrix (only using first 200 genes)
beerX <- t(as.matrix(beer.exprs))
beerX <- beerX[,1:200]
colnames(beerX) <- make.names(colnames(beerX),unique=TRUE)
##scale everything for testing.
beerX <- scale(beerX)
##define training and test sets:
set.seed(4)
allind <- 1:nrow(beerX)
learnind <- sample(allind,size=round(length(allind)*9/10))
testind <- allind[-learnind]
##fit a model with penalty equal to 10:
fit.unicox <- customSurv(X=beerX,y=beerY, learnind=learnind, lambda=10,customSurvModel=customUniCox)
##linear risk score
predict(fit.unicox@model,newdata=beerX[testind,],type='lp')
##survival probabilities at each uncensored time
mytimegrid <- sort(c(0.0,beerY[beerY[,2]==1,1]))
SurvivalProbs <- predict(fit.unicox,beerX[testind,],type='SurvivalProbs',timegrid=mytimegrid,gbm=TRUE)@SurvivalProbs
plot(SurvivalProbs,xlab="Time", ylab="Probability of Survival")

}