dab: Find bandwidth that minimizes cross validation error

In anneae/cwcens: Estimation for Component-Wise Censored Data from an Illness-Death Model

Description

This function searches a grid to find the bandwidth that minimizes mean squared leave-one-out cross validation error. The procedure uses a first-order kernel, and thus yields a bandwidth on the order of n^(-1/3). We recommend using a bandwidth on the order of n^(-7/24) in the kernel estimator, which can be obtained by multiplying the bandwidth returned by the dab function by n^(1/24).

Usage

dab(dat, lower, upper, tau, tau2, length.out = 5000)

Arguments

dat	a dataframe with one row per individual with the variables t1-tm and x1-xm where m is the number of visits and ti and xi are the time and status at each visit. xi=1 represents alive and event free (state 1 in a multistate illness-death model), and xi=2 represents alive with event.
lower	the lower end of the interval that will be searched
upper	the upper end of the interval that will be searched
tau	the maximum visit time that will be used in the cross-validation
tau2	the maximum time that individuals were at risk for a visit, based on the study design
length.out	the number of grid points in the interval [lower, upper] that will be considered. Default is 5000.

Value

the bandwidth minimizes cross-validation error. We recommend multiplying by n^(1/24) to get a bandwidth on the order of n^(-7/24) for the kernel estiamtor.

Examples

mydat <- simdat(50, scale12=1/.0008, scale13=1/.0002, scale23=1/.0016,
vital.lfu=c(30.4*36, 30.4*48),
visit.schedule = 30.4*c(6, 12, 18, 24, 30, 36, 42, 48), scatter.sd=10)
dab(mydat, lower = 30.4, upper = 36*30.4, tau = 30.4*48, tau2=30.4*48,length.out = 100)

anneae/cwcens documentation built on Aug. 14, 2021, 7:20 p.m.