aispu: Adaptive Interaction Sum of Powered Score (aiSPU) Tests
In ChongWu-Biostat/aispu: Adaptive Interaction Sum of Powered Score Tests

Description Usage Arguments Value Author(s) References Examples

View source: R/aispu.R

It returns p-values of the iSPU tests and aiSPU test.

aispu(Y, X, cov = NULL, cov2, pow = c(1:6, Inf), model = c("gaussian",
  "binomial"), n.perm = 1000, penalty = c("tlp", "lasso", "ridge", "net",
  "mcp", "SCAD"), tau = 0.1, standardize = FALSE, dfmax = 1000,
  pmax = 1000, resample = c("asy", "asy-boot", "boot"),bandwidth = 3)

`Y`	Response or phenotype data. It can be either a binary trait or a quantitative trait. A vector with length n (number of subjects).
`X`	Variables of interest; each row for a subject, and each column for predictor).A matrix with dimension n by p.
`cov`	Ancillary covariates without penalization, such as age, gender, and etc.
`cov2`	High-dimensional covariates with penalization.
`pow`	Power used in iSPU test. A vector of the powers.
`model`	Use "gaussian" for a quantitative trait, and use "binomial" for a binary trait.
`n.perm`	The number of boostrap replications
`penalty`	Penalty for cov2. The default is truncated lasso (tlp). We recommend using tlp in real data applications.
`tau`	Tunning parameters for tlp.
`standardize`	Logical flag for x variable standardization, prior to fitting the model sequence. The coefficients are always returned on the original scale. Default is standardize=TRUE. If variables are in the same units already, you might not wish to standardize.
`dfmax`	Limit the maximum number of variables in the model. Useful when the dimension of cov2 is high.
`pmax`	Limit the maximum number of variables ever to be nonzero.
`resample`	Methods for calculating p-values. The default is the fully asymptotics-based method ("asy"). ‘asy-boot’ stands for asymptotics-based method but using parametric bootstrap to calculate the mean and variance of the test statistics. ‘boot’ stands for the fully bootstrap-based method.
`bandwidth`	Optimal bandwidth for ‘asy’ method.

A list object, Ts : test statistics for the iSPU tests (in the order of the specified pow) and finally for the aSPU test. pvs : p-values for the iSPU and aiSPU tests.

Chong Wu and Wei Pan

Wu, C., Xu, G., Shen, X., & Pan, W. (2020+). A Regularization-Based Adaptive Test for High-Dimensional Generalized Linear Models, Submitted.

# Generate the data (codes for the simulations in the manuscript)
n = 30
signal.r = 0
nInformative = 3
p = 40
seed = 1
s = 0.01
non.zero = floor((p/2) * s)
alpha = c(rep(0,p/2 - non.zero), runif(non.zero,-signal.r,signal.r))
beta = c(rep(2,nInformative), rep(0,(p/2- 3)), alpha)
dat = sim_data(seed, n = n, p = p, beta = beta)

X = dat$X
Y = dat$Y

cov = NULL
X.tmp = X
cov2 = X.tmp[,1:(p/2)]
X = X.tmp[,(p/2 + 1):p]
aispu(Y, X,cov = NULL, cov2, pow = c(1:6, Inf), 
model= "gaussian",penalty = "tlp", n.perm = 10,resample = "boot")

#aispu(Y, X,cov = NULL, cov2, pow = c(1:6, Inf), 
#model= "gaussian",penalty = "tlp", n.perm = 10,resample = "asy")

#aispu(Y, X,cov = NULL, cov2, pow = c(1:6, Inf), 
#model= "gaussian",penalty = "tlp", n.perm = 10,resample = "asy-boot")