CBMAT: CBMAT

In julstpierre/CBMAT: Copula Based Multivariate Association Test for bivariate mixed phenotypes

Description

The main function for conduction Copula Based Multivariate Association Test.

Usage

CBMAT(
  y1 = NULL,
  fam1 = NULL,
  y2 = NULL,
  fam2 = NULL,
  x = cbind(rep(1, length(y1))),
  G = NULL,
  copfit = c("Gaussian", "Clayton", "Frank", "Gumbel"),
  weight = FALSE,
  weight.para1 = 1,
  weight.para2 = 25,
  pval.method = "min"
)

Arguments

y1	vector designates the first phenotype (one row-entry per individual), of length n.
fam1	a description of the error distribution and link function to be used in the marginal model of the first phenotype. Can be one of the following: "binomial(link=probit)" "gaussian()" "Gamma(link=log)" "Student(link=)"
y2	vector designates the second phenotype (one row-entry per individual), of length n.
fam2	a description of the error distribution and link function to be used in the marginal model of the second phenotype. Can be one of the following: "gaussian()" "Gamma(link=log)" "Student(link=)"
x	matrix of covariates including intercept (dimension:n \times k, with k the number of covariates), Default: cbind(rep(1, length(y1)))
G	matrix of SNPs (dimension:n \times p, with p the number of SNPs)
copfit	character, identifies the copula(s) to use for modeling phenotypes dependence. Can be any one of the following: "Gaussian", Gaussian copula "Clayton", Clayton copula "Gumbel", Gumbel copula "Frank", Frank copula , Default: c("Gaussian", "Clayton", "Frank", "Gumbel")
weight	logical, should weights be used to increase power for rare variants, Default: FALSE
weight.para1	alpha parameter of beta distribution used to simulate weights, Default: 1
weight.para2	beta parameter of beta distribution used to simulate weights, Default: 25
pval.method	character, which method to use to calculate p-value. Can be one of the following: "min", optimal p-value (default) "Fisher", Fisher's method "MFKM", MFKM method

Details

When "weight=TRUE", a weighted test is used. The weight for each SNP is a beta fuction of the corresponding minor allele frequency (MAF). There are two parameters, weight.para1 and weight.para2, for beta function. For example, when weight.para1=weight.para2=0.5, the corresponding weight is 1/sqrt(MAF*(1-MAF)); when weight.para=1 and weight.para=25, the corresponding weight is the one suggested by SKAT.

Value

A list containing results of the association test. The output list contains the following results:

• p.value: the p-value of the region-based association test

• gamma.y1: the intercept and covariates estimates from the marginal model of the first phenotype

• gamma.y2: the intercept and covariates estimates from the marginal model of the second phenotype

• alpha: the estimated phenotypic dependence parameter, as a parameter of the used copula

• tau: Kendall's tau, which is a function of the estimated phenotypic dependence parameter alpha

• cop: based on the AIC criterion calculated based on the copula-model

Author(s)

Julien St-Pierre and Karim Oualkacha

julstpierre/CBMAT documentation built on Aug. 7, 2021, 9:31 p.m.