How to calculate brier score for a given set of probabilities in R for survival outcome?

Question

I have a survival dataset where time and status variable are given along with some survival probabilities at (say) time = 12 , calculated based on some indices for each individual in the dataset. Now, I would like to calculate the brier scores for each index in the dataset and choose the best one. Here is my sample code:

set.seed(99)
library(dplyr)
library(survival)
library(prodlim) 
library(pec)   

# simulated survival data data
data <- SimSurv(100)

# add two incides
data$index1<-runif(100, 0.4, 0.9)
data$index2<-runif(100, 0.5, 1)

I tried the following using pec package in R .

# let's try for one index
models<-as.matrix(data$index1)

# try pec package, which provides brier score
PredError <- pec(object=models,
             formula = Surv(time, status)~1, cens.model="marginal",
             data=data, verbose=F, maxtime=200, times = 12)

But I got the following error message:

Error in predictSurvProb.matrix(object = c(0.596092602680437, 0.675279439869337,  :

Prediction matrix has wrong dimensions:
Requested newdata x times: 100 x 102
Provided prediction matrix: 100 x 1

Anyone has any suggestion on how to resolve the issue?

Answer 1

While using matrix as object times argument is adjusted inside the function so if you change the times argument to zero and exact to FALSE, code runs successfully.

models <- as.matrix(data$index1)

PredError <- pec(object = models, formula = Surv(time, status) ~ 1, cens.model = "marginal", data = data, verbose = T, maxtime = 200, times = 0, exact = FALSE)

Also if you change the object to other type except matrix then the code works fine.

For example

models <- list("Cox.X1" = coxph(Surv(time, status) ~ index1, data = data, x = TRUE, y = TRUE))

PredError <- pec(object = models, formula = Surv(time, status) ~ 1, cens.model = "marginal", data = data, verbose = T, maxtime = 200, times = 12)

For details regarding the object argument please refer below statement:

A named list of prediction models, where allowed entries are (1) R-objects for which a predictSurvProb method exists (see details), (2) a call that evaluates to such an R-object (see examples), (3) a matrix with predicted probabilities having as many rows as data and as many columns as times. For cross-validation all objects in this list must include their call.