WEE.linear: WEE linear regression

In WEE: Weighted Estimated Equation (WEE) Approaches in Genetic Case-Control Studies

Description

Returns an object of class "WEE.linear" that is generated by linear regression with WEE approach for continuous secondary traits in genetic case-control studies.

Usage

WEE.linear(formula, D, data, pd_pop, boot = 0, ...)

Arguments

formula	the secondary trait given SNPs and covariates. e.g. y~x+z
D	primary disease (case-control status)
data	dataset with real observation.
pd_pop	the population disease prevelance of primary disease.
boot	number of bootstrap samples. (boot=0 by default)
...	optional arguments to be passed through to lm.

Value

Coefficients	Point estimates
StdErr	Bootstrap standard errors, returned if boot > 0
Chisq	Chi-squared test statistics, returned if boot > 0
p.value	p-values, returned if boot > 0
Covariance	Covariance matrix, returned if boot > 0

Warning

If boot = 0, point estimates are plotted. If boot > 0, boostrap standard errors, chisquare test statistics, p-values, and covariance matrix are also returned. Optional arguments from lm can be passed to this function, but arguments 'subset' and 'weights' should be used with caution.

References

Xiaoyu Song, Iuliana Ionita-Laza, Mengling Liu, Joan Reibman, Ying Wei (2016). A General and Robust Framework for Secondary Traits Analysis. Genetics, vol. 202 no. 4 1329-1343; DOI: 10.1534/genetics.115.181073

Examples

## Generate simulated data
# set population size as 500000
n = 500000

# set parameters
beta = c(0.2, 0.1) # P(Y|X,Z)
gamma = c(0.3, log(2), log(2)) #P(D|X,Y,Z)

# generate the genetic variant X
x = rbinom(n,size=2,prob=0.3)

# generate the standardized continuous covariate Z correlated with X
z = rnorm(n, mean=0.5*x-0.3, sd=1)

# generate the continuous secondary trait Y
y = 1+beta[1]*x+beta[2]*z+rnorm(n)

# generate the primary disease D
alpha = -3.62
pd = exp(alpha + x*gamma[1] + y*log(2)+ z*log(2)) /
         (1+exp(alpha+ x*gamma[1] + y*log(2)+ z*log(2)))
d = rbinom(n,size=1,prob=pd)

# form population data set
dat=as.data.frame(cbind(d, y, z, x))

# generate sample dataset with 200 cases and 200 controls
dat_cases = dat[which(dat$d==1),]
dat_controls= dat[which(dat$d==0),]
dat_cases_sample = dat_cases[sample(sum(dat$d==1),
                             200, replace=FALSE),]
dat_controls_sample = dat_controls[sample(sum(dat$d==0),
                                  200, replace=FALSE),]
  
dat_linear=rbind(dat_cases_sample,dat_controls_sample)
colnames(dat_linear)=c("D", "y", "z","x")
D = dat_linear$D # Disease status
pD = sum(dat$d == 1)/500000 # Population disease prevalence

## WEE linear regresssion
WEE.linear(y ~ x + z, D, 
           data = dat_linear, pD)   

WEE.linear(y ~ x + z, D, 
           data = dat_linear, pD, boot = 500) 

WEE documentation built on May 2, 2019, 6:43 a.m.