epsCC.test: Score test EPS-CC

In theabjorn/extremesampling: Asymptotic tests and model fitting for extreme phenotype sampling

Description

epsCC.test performs a score test for common genetic variants under the EPS complete-case design

Usage

epsCC.test(nullmodel, xg, l, u)

Arguments

nullmodel	an object of class formula, that describes the linear regression model under the null hypothesis
xg	a matrix of genetic variants to be tested against the null
l	cutoff for lower extreme, can be sample-specific or general
u	cutoff for upper extreme, can be sample-specific or general

Details

The nullmodel formula object is of the type y~xe, which describes a regression model, y=a+be*xe+e assuming a normal distribution for the residuals (e). The covariate xe is a non-genetic/environmental covariate (optional). The variables are taken from the environment that the function is called from. The null hypothesis bg=0 is tested for the model y=a+be*xe+bg*xg+e. The covariate xg is one or more genetic markers.

Variables are only available for individuals with high and low values of the phenotype y; (y < l or y > u), and potentialy some randomly sampled individuals. The cut-offs l and u that specify the sampling must be given in the cutoffs argument.

Value

epsCC.test returns for each genetic variant:

statistic	the score test statistic
p.value	the P-value

Examples

N = 1000
# Generate environmental covariates
xe1 = rbinom(N,1,0.5); xe2 = rnorm(N,2,1)
# Generate genetic covariates (common variants)
cv1 = rbinom(N,2,0.2); cv2 = rbinom(N,2,0.2)
# Generate phenotype
y = rnorm(N, mean = 1 + 2*xe1 + 3*xe2 + 0.5*cv1 + 0.1*cv2,2)
# Define extremes
u = quantile(y,probs = 3/4,na.rm=TRUE); l = quantile(y,probs = 1/4,na.rm=TRUE)
extreme = (y < l) | (y >= u)
cv1[!extreme] = NA; cv2[!extreme] = NA;

# Complete case data set
xe_CC = cbind(xe1[extreme], xe2[extreme])
xg_CC = cbind(cv1[extreme], cv2[extreme])
y_CC = y[extreme]

epsCC.test(y_CC ~ xe_CC,xg = xg_CC,l,u)

theabjorn/extremesampling documentation built on May 31, 2019, 9:10 a.m.