getSplitMethod: Input for data splitting algorithms

In riskRegression: Risk Regression Models and Prediction Scores for Survival Analysis with Competing Risks

Input for data splitting algorithms

Description

Parse hyperparameters for data splitting algorithm

Usage

getSplitMethod(split.method, B, N, M, seed)

Arguments

split.method	A character string specifying the algorithm for data splitting: "loob" leave one out bootstrap "bootcv" bootstrap cross validation "cv5" 5-fold cross validation "loocv" leave one out cross validation aka N-1 fold cross validation "632plus" Efron's .632+ bootstrap
B	Number of repetitions of bootstrap or k-fold cross-validation
N	Sample size
M	Subsample size. Default is N (no subsampling).
seed	Integer passed to set.seed. If not given or NA no seed is set.

Value

A list with the following elements:

• split.methodName: the print name of the algorithm

• split.method: the internal name of the algorithm

• index: the index for data splitting. For bootstrap splitting this is a matrix with B columns and M rows identifying the in-bag subjects. For k-fold cross-validation this is a matrix with B columns identifying the membership to the k groups.

• k: the k of k-fold cross-validation

• N: the sample size

• M: the subsample size

Author(s)

Thomas A. Gerds <[e-mail]>

See Also

Score

Examples

# 3-fold crossvalidation
getSplitMethod("cv3",B=4,N=37)

# bootstrap with replacement
getSplitMethod("loob",B=4,N=37)

# bootstrap without replacement
getSplitMethod("loob",B=4,N=37,M=20)

riskRegression documentation built on May 29, 2024, 10:59 a.m.