ICSKAT (version 0.3.0)

ICSKATO_bootstrap: ICSKATO_bootstrap.R

Description

The version of ICSKATO to run when bootstrapping to match kurtosis.

Usage

ICSKATO_bootstrap(
  icskatOut,
  B,
  intervalProbs,
  allVisits,
  quant_r,
  seed = NULL,
  null_fit,
  gMat,
  xMat,
  fitAgain,
  checkpoint = FALSE,
  downsample = 1,
  rhoVec
)

Value

A list with the elements:

kurtQvec

Vector of bootstrapped excess kurtosis of each Qrho.

varQvec

Vector of bootstrapped variance of each Qrho.

meanQvec

Vector of bootstrapped mean of each Qrho.

kurtKappa

Bootstrapped kurtosis of kappa term without zeta.

kurtKappaAll

Bootstrapped kurtosis of full kappa term with zeta.

varKappaAll

Bootstrapped variance of full kappa term with zeta.

meanKappaAll

Bootstrapped mean of full kappa term with zeta.

bootDF

Matrix with B rows containing all the bootstrapped quantities over all iterations.

QrhoBoot

Matrix with B rows containing all the bootstrapped Qrho values, one column for each rho.

listDS

A list containing all of the other elements in this return list, except using the downsampled iterations.

nonNA

Number of bootstraps that did not result in NA (and thus were not removed).

Arguments

icskatOut

The output list from ICSKAT().

B

Number of bootstrap replications.

intervalProbs

n*(s+1) matrix where n is number of subjects and s is the number of visits possible, probability of falling in each interval.

allVisits

n*s matrix with all the visit times for each subject.

quant_r

Quantiles of time from make_IC_dmats, to keep them constant through bootstrapping.

seed

Seed to start the bootstrapping.

null_fit

The null fit output from ICSKAT_fit_null.

gMat

Genotype matrix used in original test.

xMat

n*p matrix of non-genetic covariates.

fitAgain

Boolean, whether to fit the null model again in each bootstrap.

checkpoint

Boolean, whether to print every time 100 bootstraps finish.

downsample

A number in (0, 1], will use this fraction of the bootstrap iterations to try running the test with fewer bootstraps.

rhoVec

Vector of rhos to search over in SKATO.

Examples

set.seed(2)
gMat <- matrix(data=rbinom(n=2000, size=2, prob=0.3), nrow=100)
xMat <- matrix(data=rnorm(200), nrow=100)
bhFunInv <- function(x) {x}
obsTimes <- 1:5
etaVec <- rep(0, 100)
outcomeDat <- gen_IC_data(bhFunInv = bhFunInv, obsTimes = obsTimes, windowHalf = 0.1,
probMiss = 0.1, etaVec = etaVec)
lt <- outcomeDat$leftTimes
rt <- outcomeDat$rightTimes
tpos_ind <- as.numeric(lt > 0)
obs_ind <- as.numeric(rt != Inf)
dmats <- make_IC_dmat(xMat, lt, rt, obs_ind, tpos_ind)
nullFit <- ICSKAT_fit_null(init_beta = rep(0, 5), left_dmat = dmats$left_dmat,
right_dmat=dmats$right_dmat, obs_ind = obs_ind, tpos_ind = tpos_ind,
lt = lt, rt = rt)
icskatOut <- ICskat(left_dmat = dmats$left_dmat, right_dmat=dmats$right_dmat,
lt = lt, rt = rt, obs_ind = obs_ind, tpos_ind = tpos_ind, gMat = gMat,
null_beta = nullFit$beta_fit, Itt = nullFit$Itt)
intervalProbOutput <- construct_interval_probs(allTimes = outcomeDat$allVisits,
dmats = dmats, nullBeta = nullFit$beta_fit, p = ncol(xMat), nKnots=1)
ICSKATO_bootstrap(icskatOut = icSkatOut, B = 100, intervalProbs = intervalProbOutput$probMat,
allVisits = intervalProbOutput$allTimesFilled, quant_r = dmats$quant_r, seed = 0,
null_fit = nullFit, gMat = gMat, xMat, fitAgain = TRUE,
rhoVec=c(0, 0.01, 0.04, 0.09, 0.25, 0.5, 1))