Package 'survivalisttoo'

Model-Agnostic Survival Analysis

Description

Model-agnostic survival analysis: using any Machine learning algorithm for doing survival analysis.

Package Content

Index of help topics:

cox_gradient_boost      Cox Gradient Boosting Model
predict.CoxGradientBoost
                        Predict from a CoxGradientBoost model
print.CoxGradientBoost
                        Print a CoxGradientBoost Object
survivalisttoo-package
                        Model-Agnostic Survival Analysis

Maintainer

T. Moudiki <[email protected]>

Author(s)

T. Moudiki [aut, cre]

---

Cox Gradient Boosting Model

Description

Fit a model-agnostic Cox-gradient boosting survival model

Usage

cox_gradient_boost(
  X,
  time,
  event,
  regr_fun,
  M = 100L,
  nu = 0.1,
  show_progress = TRUE,
  ...
)

Arguments

X	numeric matrix of covariates (n x p)
time	numeric vector of observed times
event	integer/logical vector (1 = event, 0 = censored)
regr_fun	function(X, y, ...) -> model with predict() method
M	number of boosting iterations (default 100)
nu	learning rate / shrinkage (default 0.1)
show_progress	logical; print a progress bar? (default TRUE)
...	extra arguments forwarded to regr_fun at every iteration

Value

object of class "CoxGradientBoost"

Examples

require(glmnet)
require(survival)

data(ovarian)

set.seed(42)
idx_train <- sample(nrow(ovarian), floor(0.75 * nrow(ovarian)))
df <- ovarian
train <- df[idx_train, ]; test <- df[-idx_train, ]

regr_lm <- function(X, y, ...) lm(y ~ ., data = data.frame(X, y = y))

fit_boost_lm <- survivalisttoo::cox_gradient_boost(train, train$futime, 
train$fustat, regr_lm)

y_test   <- Surv(test$futime, test$fustat)
(ci_blm <- glmnet::Cindex(predict(fit_boost_lm, test), y_test)) # C-index

---

Predict from a CoxGradientBoost model

Description

Predict from a CoxGradientBoost model

Usage

## S3 method for class 'CoxGradientBoost'
predict(object, newdata, type = "lp", times = NULL, M = NULL, ...)

Arguments

object	a CoxGradientBoost object
newdata	numeric matrix or data.frame of new covariates
type	"lp" (default), "survival", or "cumhaz"
times	time grid for type != "lp"; NULL -> training event times
M	use only the first M base learners (early stopping probe); NULL -> all M (default)
...	unused

Value

type = "lp" -> numeric vector (length n_new) type = "survival" -> matrix (n_new x length(times)) type = "cumhaz" -> matrix (n_new x length(times))

---

Print a CoxGradientBoost Object

Description

Displays a concise summary of a fitted CoxGradientBoost model, including the base learner, number of boosting iterations, learning rate, and basic information about the training data.

Usage

## S3 method for class 'CoxGradientBoost'
print(x, ...)

Arguments

x	An object of class "CoxGradientBoost".
...	Further arguments passed to or from other methods (currently unused).

Details

This method prints key characteristics of the fitted model:

• The base learner used in boosting

• The number of boosting iterations ($M$)

• The learning rate ($\nu$)

• The number of observations

• The number of observed events

• The number of unique event times

Value

The input object x, invisibly.

See Also

cox_gradient_boost, predict.CoxGradientBoost

---