kkm: Kernel Weighted Kaplan-Meier model for survival

In linxihui/sml: Machine Learning for Survival Data

Description

Kernel Weighted Kaplan-Meier model for survival

Usage

kkm(x, ...)

## S3 method for class 'formula'
kkm(formula, data, xtest = NULL, ytest = NULL, ...)

## Default S3 method:
kkm(x, y, xtest = NULL, ytest = NULL,
  return.train.prediction = FALSE || is.null(xtest), scaled = TRUE,
  k = min(30, nrow(x)), times = y[as.logical(y[, 2]), 1],
  type = c("kaplan-meier", "nelson-aalen"), kernel = "rbfdot",
  kpar = "automatic")

Arguments

x	Design matrix (NO intercept)
...	Further argument passed to internal functions
formula	Model formula
data	Data frame
xtest	If the 'formula-data' format used, 'xtest' is a data frame of the test set. If the 'x-y' method called, 'xtest' is the design matrix of the test set
ytest	(optional) Survival outcome for testset
y	Reponse vector of 'Surv' object
scaled	Logical value indicating if to standardize x, y
k	Number of nearest neighbour used. Default to 'nrow(x)' which seems the best
times	Times to evaluate survival probabilities. Currently no used. All unique event times in training are used
type	Type of estimator, either 'Kaplan-meier' or 'nelson-aalen'
kernel	String or 'kernel' object (see kernlab::gausspr)
kpar	A list of Kernel parameters or 'automatic' if a radial kernel specified

Value

A 'kkm' object

Examples

library(survival);
data(pbc, package = 'randomForestSRC');
pbc <- na.omit(pbc);
i.tr <- sample(nrow(pbc), 100);
kkm.pred <- kkm(Surv(days, status) ~., data = pbc[i.tr, ], xtest = pbc[-i.tr, ], kernel = 'laplacedot', kpar = list(sigma = 0.1));
# concordance index if using 30th event time
survConcordance(Surv(days, status) ~ I(1 - kkm.pred$test.predicted.survival[, 30]), data = pbc[-i.tr, ])$concordance

plot(kkm.pred, subset = sample(length(i.tr), 10), lwd = 2)

linxihui/sml documentation built on May 21, 2019, 6:39 a.m.