modelFit: Model fit for the training set
In Yinghao-Pan/reinforcedPred: Reinforced Risk Prediction with Budget Constraint

Description Usage Arguments Value Examples

modelFit outputs the FPCA (functional principal component analysis) decomposition and the parameter estimates of the functional generalized linear model at each time grid.

1	modelFit(Y, X, Z, startT, link, pve, nbasis, weight)

`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.
`link`	The link function used in functional generalized linear models, e.g. "logit", "probit".
`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.

`list_fpcaFit`	FPCA decomposition at each time grid from startT to the end.
`list_paraEst`	Parameter estimates at each time grid from startT to the end.

library(reinforcedPred)

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

# analysis starts
startT <- 55
link <- "probit"
weight <- rep(1, length(Y))

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

# obtained parameter estimates and FPCA decompositions
list_paraEst <- result$list_paraEst
list_fpcaFit <- result$list_fpcaFit