WEE.logistic: WEE logistic regression

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

Description

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

Usage

WEE.logistic(formula, D, data, pd_pop, iter = 5, 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.
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 glm.

Value

Coefficients	Point estimates
Oddsratio	The exponentiated coefficients, namely the odds ratio associated with the corresponding covariate
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 glm 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 binary secondary trait Y
py = exp(-1+beta[1]*x+beta[2]*z)/
         (1+exp(-1+beta[1]*x+beta[2]*z))
y = rbinom(n,1, py)

# generate the primary disease D
# (alpha changes to make sure the disease prevalence = 0.1 )
alpha = -2.88
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 the population dataset
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_logistic = rbind(dat_cases_sample,dat_controls_sample)	
colnames(dat_logistic) = c("D", "y", "z","x")
D = dat_logistic$D # Disease status
pD = sum(dat$d==1)/500000 # Population disease prevalence

## WEE logsitic regression
WEE.logistic(y ~ x + z, D, 
             data = dat_logistic, pD)	  

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

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