modelFit_VS: Model fit for the training set, variable selection version

Description Usage Arguments Value Examples

Description

modelFit_VS outputs the FPCA (functional principal component analysis) decomposition and the elastic net logistic regression at each time grid. This function is used when the baseline covariates Z are high-dimensional.

Usage

1
modelFit_VS(Y, X, Z, startT, pve, nbasis, weight)

Arguments

Y

The outcome variable, vector of length n, taking values in {1, 0, NA}, where 1 = disease, 0 = not, NA = missing.

X

Observed longitudinal biomarker, matrix of n by nTotal, where nTotal denotes the total number of time grids. Missing values are denoted by NA.

Z

Other baseline covariates.

startT

Time of the first prediction, denoted by t_1 in the manuscript. For instance, if the time grids are {0,1/60,2/60,...,1}, then startT = 25 means that the first prediction is made at t = 24/60.

pve

Proportion of variance explained in FPCA.

nbasis

Number of B-spline basis functions needed for estimation of the mean function and smoothing of covariance.

weight

Weight for each individual.

Value

list_fpcaFit

FPCA decomposition at each time grid from startT to the end.

list_cvfit

Elastic net logistic regression at each time grid from startT to the end.

Examples

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
library(reinforcedPred)

# take the example training data (high dimensional Z) from the reinforcedPred package
# see documentation for details about the data set train_data_mulZ
Y <- as.numeric(train_data_mulZ$Y)
tildeX.missing <- as.matrix(train_data_mulZ[,2:62])
Z <- as.matrix(train_data_mulZ[,63:dim(train_data_mulZ)[2]])

# analysis starts
startT <- 25
weight <- rep(1, length(Y))

result <- modelFit_VS(Y, tildeX.missing, Z, startT, pve = 0.99, nbasis = 10, weight)

# obtained elastic net logistic regression fit and FPCA decompositions
list_cvfit <- result$list_cvfit
list_fpcaFit <- result$list_fpcaFit

Yinghao-Pan/reinforcedPred documentation built on May 10, 2019, 8:24 a.m.