WEE.quantile: WEE quantile regression

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

Description

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

Usage

WEE.quantile(formula, D, data, pd_pop, tau, iter = 5, boot = 0, ...)

Arguments

formula	The secondary trait given SNPs and covariates. e.g. y~x+z
D	Primary disease (case-control status), must be specified.
data	Dataset with real observation.
pd_pop	The population disease prevelance of primary disease.
tau	The quantile level to be estimated. Multiple taus can be chosen.
iter	Number of generating pseudo observations. (iter=10 by default)
boot	Number of bootstrape samples. (boot=0 by default)
...	Optional arguments to be passed through to rq.

Details

The quantile regression package "quantreg" is required before calling this function

Value

Coefficients	Point estimates
StdErr	Bootstrap standard errors, returned if boot > 0
Wald	Wald 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, Wald test statistics, p-values, and covariance matrix are also returned. Optional arguments from rq can be passed to this function, but arguments 'subset' and 'weights' should be used with caution.

References

[1] Ying Wei, Xiaoyu Song, Mengling Liu, Iuliana Ionita-Laza and Joan Reibman (2016). Quantile Regression in the Secondary Analysis of Case Control Data. Journal of the American Statistical Association, 111:513, 344-354; DOI: 10.1080/01621459.2015.1008101

[2] 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(0.12, 0.1) # P(Y|X,Z)
gamma = c(-4, log(1.5), log(1.5), log(2) ) #P(D|X,Y,Z)

# generate the genetic variant X
x = rbinom(n,size=2,prob=0.3)
  
# generate the continuous covariate Z
z = rnorm(n)
  
# generate the continuous secondary trait Y
y= 1 + beta[1]*x + beta[2]*z + (1+0.02*x)*rnorm(n)

# generate disease status D
p = exp(gamma[1]+x*gamma[2]+z*gamma[3]+y*gamma[4])/
(1+exp(gamma[1]+x*gamma[2]+z*gamma[3]+y*gamma[4]))
d = rbinom(n,size=1,prob=p)

# form population data dataset
dat = as.data.frame(cbind(x,y,z,d))
colnames(dat) = c("x","y","z","d")

# 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_quantile = as.data.frame(rbind(dat_cases_sample,
                                   dat_controls_sample))
colnames(dat_quantile) = c("x","y","z","D")
D = dat_quantile$D # Disease status
pd = sum(d==1)/n # population disease prevalence

# WEE quantile regressions:
WEE.quantile(y ~ x, D, tau = 0.5, 
             data = dat_quantile, pd_pop = pd)	  				

WEE.quantile(y ~ x + z, D, tau = 1:9/10, 
             data = dat_quantile, pd_pop = pd, boot = 500)

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