SurvRank (version 0.1)

CVrankSurv_fct: Main function of SurvRank.

Description

Main input function for SurvRank.

Usage

CVrankSurv_fct(data, t.times, cv.out, cv.in, fs.method = "lasso.rank", nr.var = 10, sd1 = 0.95, ncl = 1, weig.t = T, n1 = 0.1, c.time = 10, ...)

Arguments

data

input of data as a list in the format: list.name$x data.frame of covariates. list.name$y response as a survival object, derived from Surv.

t.times

number of times the cross-validation should be repeated

cv.out

number of folds in outer cross validation loop (for estimation of the predictive accuracy)

cv.in

number of folds in inner cross validation loop (for model selection on the training set)

fs.method

Defaults to "lasso.rank". Ranking method to be applied. One of c("lasso.rank","conc.rank","rf.rank","boost.rank","cox.rank","rpart.rank","randcox.rank","wang.rank")

nr.var

Number of variables up to which stepwise selection should be carried out. Has to be smaller than n number of observations.

sd1

factor to which sparser solutions should be chosen. Not maximum Survival AUC in inner loop is used in stepwise selection, instead max(survAUC)*sd1 leading to sparser solutions

ncl

Defaults to 1. Number of clusters for parallel execution.

weig.t

Defaults to TRUE. Should a weighting of features be performed.

n1

used in weighting function if weig.t=T. Find details in weighting_fct

c.time

as defined in package survAUC time; a positive number restricting the upper limit of the time range under consideration.

...

arguments that can be passed to underlying functions, not used now

Value

Output of the CVrankSurv_fct, basically a list containing the following elements

method

ranking method

accuracy$ranking

full ranking of all model estimations

accuracy$pred.in

averaged inner AUCs of stepwise selection

accuracy$pred.out

predictions of testset

accuracy$used.rank

only used features according to stepwise selection

accuracy$used.rank1se

only used features according to stepwise selection with factor sd1

accuracy$auc.out

matrix of dimension cv.out times t.times of survival AUCs.

accuracy$auc.out1se

matrix of dimension cv.out times t.times of survival AUCs with factor sd1.

rank.mat

matrix of ranks per feature. If not selected, it is set to number of features.

out.mat

0/1 matrix for features selected

out.mat1se

0/1 matrix for features selected with factor sd1 application

top1se

unweighted toplist with factor sd1

toplist

unweighted toplist

weighted

weighted toplist with applied weighting function

rank

toplist of ranked features according to ranks

Details

details to follow

Examples

## Simulating a survival data set
N=100; p=10; n=4
x=data.frame(matrix(rnorm(N*p),nrow=N,p))
beta=rnorm(n)
mx=matrix(rnorm(N*n),N,n)
fx=mx[,seq(n)]%*%beta/3
hx=exp(fx)
ty=rexp(N,hx)
tcens=1-rbinom(n=N,prob=.3,size=1)
y=Surv(ty,tcens)
data=list()
data$x<-x; data$y<-y
## Ranking the features according to their significance in the univariate cox models
out.cox<-CVrankSurv_fct(data,2,3,3,fs.method="cox.rank")
## Ranking the features according to the LASSO algorithm
## Not run: 
# out.lasso<-CVrankSurv_fct(data,2,5,5,fs.method="lasso.rank")## End(Not run)