validate: validate

In AxelitoMartin/OncoCast: Ensemble learning survival prediction with improved interface.

Description

This function let's user validate an OncoCast model generated in a new cohort which has it's own survival data.

Usage

validate(
  OC_object,
  Results,
  in.data,
  formula,
  out.ties = FALSE,
  limit = NULL,
  plot.cuts = TRUE,
  ...
)

Arguments

OC_object	Output of the OncoCast function.
Results	A getResults_OC object that you wish to validate
in.data	New data set containing the information of the incoming patients. Should be a dataframe with patients as rows and features as columns.
formula	the formula to be used for validation
out.ties	phcpe argument to calculate the concordance index. If out.ties is set to FALSE, pairs of observations tied on covariates will be used to calculate the CPE. Otherwise, they will not be used.
limit	Optional numerical argument to set a time limit on the KM plot
plot.cuts	Boolean specifying if lines should be added to risk histogram to indicate where the cuts were performed
...	futher arguments applicable in ggsurvplot

Value

data.out : The data frame inputted in the function with an additional column giving the predicted risk score of the incoming patients.

RiskHist : A histogram of the distribution of the risk scores of patients in the given dataset.

Examples

library(OncoCast)
test <- OncoCast(data=survData,formula = Surv(time,status)~.,
method = "LASSO",runs = 30,
save = FALSE,nonPenCol = NULL,cores =1)
results <- getResults_OC(OC_object=test$LASSO,data=survData,
cuts=c(0.2,0.4,0.6,0.8),
geneList=NULL,mut.data=TRUE)
new.data <- as.data.frame(matrix(rbinom(5*20,1,0.5),nrow=20,ncol = 5))
colnames(new.data) <- c("ImpCov1","ImpCov2","ImpCov3","ImpCov4","Cov7")
rownames(new.data) <- paste0("Incoming",1:20)

# add time component
coefs <- c(1.2,1.3,-1.5,-1,-1,rep(0,ncol(new.data)-5))
mu <- as.vector(coefs %*% t(new.data))
n = nrow(row(new.data))
time <- rexp(n,exp(mu))*12
c <- runif(n,0,as.numeric(quantile(time,0.8)))
status <- ifelse(time < c , 1,0)
time <-pmin(time,c)
new.data$time <- time
new.data$status <- status

validation <- validate(OC_object=test$LASSO,Results=results,
in.data=new.data, formula=Surv(time,status)~.)

AxelitoMartin/OncoCast documentation built on Dec. 13, 2020, 6:46 a.m.