| GEEaSPU {GEEaSPU} | R Documentation |
An adaptive sum of powered score (SPU) test for single SNP and mutiple phenotypes
Description
Tests whether a SNP is associated with multivariate outcomes; provides a series of p-values of GEE-SPU(pow) and GEE-aSPU tests.
Usage
GEEaSPU(pheno, geno, Z = NULL, model = "gaussian", corstr = "independence",
pow = c(1:8, Inf), n.perm = 1000, null.type = "perm", score.test = FALSE)
Arguments
pheno |
a numeric phenotype matrix with each row as a different individual and each column as a separate phenotype. |
geno |
a numeric vector with each element for an individual genotype. |
Z |
a numeric covariate matrix with each row as a different individual and each column as a covariated to be adjested. |
model |
a character string specifying the model of the phenotypes. Models supported are "gaussian" for a quantitative trait and "binomial" for a binary trait (default = "gaussian"). |
corstr |
a character string specifying the correlation structure of phenotypes. The following are permitted: "independence", "fixed", "stat_M_dep", "non_stat_M_dep", "exchangeable", "AR-M" and "unstructured" (default = "independence"). |
pow |
a vector of the power weight to be used at a trait level (default = c(1:8, Inf)). |
n.perm |
a numeric value of number of null statistics (default = 1000). |
null.type |
a character string specifying how to generate null statistics; "perm" is used when null statistics are generated using permutations and "sim" is used when null statistics are generated using simulations (default = "perm"). |
score.test |
a logical value indicating whether to include GEEaSPU-Score test along with GEE-Score test (default = FALSE). If TRUE, it computes p-values of GEEaSPU-Score and GEE-Score as well as GEEaSPU test. |
Details
Adaptive association tests for single SNP and multiple phenotypes using GEE.
Value
a vector of p-values from GEE-SPU(pow) tests and GEE-aSPU test.
Note
When large SNP-set (namely large gene size) or large number of phenotypes are included, the permuation based test (null.type = "perm") is recommended.
An option "binomial" model only supports the option, null.type = "sim".
Author(s)
Junghi Kim and Wei Pan
References
Kim J, Zhang Y, and Pan W (2016). "Powerful and adaptive testing for multi-trait and multi-SNP associations with GWAS and sequencing data." Genetics 203(2), 715–731.
Zhang Y, Xu Z, Shen, X and Pan W (2014). "Testing for association with multiple traits in generalized estimation equations, with application to neuroimaging data." NeuroImage, 96, 309–325.
Examples
# -- simulating phenotypes
# -- n.subjects: number of subjects
# -- n.traits: number of phenotypes
# -- Sigma: covariance matrix of phenotypes (e.g. AR(1))
set.seed(136)
n.subjects <- 100
n.traits <- 3
sigma <- 2; rho <- 0.5
Sigma0 <- diag(n.traits);
Sigma <- sigma * rho^abs(row(Sigma0) - col(Sigma0))
eS <- eigen(Sigma, symmetric = TRUE)
ev <- eS$values
X <- matrix(rnorm(n.traits * n.subjects), n.subjects)
pheno <- X %*% diag(sqrt(pmax(ev, 0)), ncol(Sigma)) %*% eS$vectors
# -- simulating genotype
geno <- rbinom(n = n.subjects, size = 2, prob = 0.2)
# -- Computing the p-value of GEEaSPU test with the permutation based method
Pvl <- GEEaSPU(pheno = pheno, geno = geno, Z = NULL, pow = c(1,2,4,Inf),
n.perm = 1000, null.type = "perm", score.test = FALSE)
# -- Each element of Pvl is a p value of GEE-SPU(pow) in order
# -- The last element of Pvl is a p value of GEE-aSPU test
Pvl
Pvl[length(Pvl)]
# > Pvl
# SPU.1 SPU.2 SPU.4 SPU.Inf aSPU
# 0.1890000 0.4070000 0.3520000 0.3040000 0.2917083
# > Pvl[length(Pvl)]
# aSPU
# 0.2917083