cv.imcid: Selection of the tuning parameters for determining the MCID...

In MCID: Estimating the Minimal Clinically Important Difference

Description

cv.imcid returns the optimal tuning parameter δ and λ selected from a given grid by using k-fold cross-validation. The tuning parameters are selected for determining the MCID at the individual level

Usage

cv.imcid(x, y, z, lamseq, delseq, k = 5, maxit = 100, tol = 0.01)

Arguments

x	a continuous variable denoting the outcome change of interest
y	a binary variable denoting the patient-reported outcome derived from the anchor question
z	a vector or matrix denoting the patient's clinical profiles
lamseq	a vector containing the candidate values for the tuning parameter λ, where λ is the coefficient of the penalty term, used for avoiding the issue of model overfitting
delseq	a vector containing the candidate values for the tuning parameter δ, where δ is used to control the difference between the 0-1 loss and the surrogate loss. We recommend selecting the possible values from the neighborhood of the standard deviation of x
k	the number of groups into which the data should be split to select the tuning parameter δ by cross-validation. Defaults to 5
maxit	the maximum number of iterations. Defaults to 100
tol	the convergence tolerance. Defaults to 0.01

Value

a list including the combinations of the selected tuning parameters and the value of the corresponding target function

Examples

n <- 500
lambdaseq <- 10 ^ seq(-3, 3, 0.1)
deltaseq <- seq(0.1, 0.3, 0.1)
a <- 0.1
b <- 0.55
c <- -0.1
d <- 0.45

set.seed(721)
p <- 0.5
y <- 2 * rbinom(n, 1, p) - 1
z <- rnorm(n, 1, 0.1)
y_1 <- which(y == 1)
y_0 <- which(y == -1)
x <- c()
x[y_1] <- a + z[y_1] * b + rnorm(length(y_1), 0, 0.1)
x[y_0] <- c + z[y_0] * d + rnorm(length(y_0), 0, 0.1)

sel <- cv.imcid(x = x, y = y, z = z, lamseq = lambdaseq,
         delseq = deltaseq, k = 5, maxit = 100, tol = 1e-02)
sel$'Selected lambda'
sel$'Selected delta'

MCID documentation built on Sept. 10, 2021, 5:07 p.m.