NelsonAalen {ePCR}

Cox-Oakes extension of the Nelson-Aalen estimates for a Cox model

Description

Implementing the heuristic Cox and Oakes extension of the Nelson-Aalen estimate for Cox model to extract individual-specific survival. Time-to-event predictions are then given at the first time point at which an individual reaches an event probability of 50

Usage

NelsonAalen(
  b,
  Xold,
  Xnew,
  events,
  time,
  tpred = 0:round(max(time, na.rm = T), 0),
  plot = FALSE
)

Arguments

b

Beta coefficients at optimal glmnet coxnet model (lambda, alpha)

Xold

Data matrix with rows as individuals (i.e. training data)

Xnew

Possible new prediction data matrix (if omitted the training data is used, or if it is a new dataset the columns should comform to the training data and new individuals be provided as rows)

events

Deaths or right-censoring per each individual (1 death 0 alive censored) for Xold

time

Times to event or censoring for Xold

tpred

The predicted time points; more tight grid gives a smoother curve

plot

Should an individualized plot be plotted to show how the cumulative survival curves behave

Value

Predicted times-to-event predictions either for the training data or an optional provided novel dataset

Note

See section 3.6 at http://data.princeton.edu/pop509/NonParametricSurvival.pdf for the reference

Author(s)

Teemu Daniel Laajala [e-mail]

Examples

data(TYKSSIMU)
library(survival)
xdat <- as.matrix(xMEDISIMU)
ydat <- yMEDISIMU[,"surv"]