gcontrol: genomic control

In gap: Genetic Analysis Package

genomic control

Description

genomic control

Usage

gcontrol(
  data,
  zeta = 1000,
  kappa = 4,
  tau2 = 1,
  epsilon = 0.01,
  ngib = 500,
  burn = 50,
  idum = 2348
)

Arguments

data	the data matrix.
zeta	program constant with default value 1000.
kappa	multiplier in prior for mean with default value 4.
tau2	multiplier in prior for variance with default value 1.
epsilon	prior probability of marker association with default value 0.01.
ngib	number of Gibbs steps, with default value 500.
burn	number of burn-ins with default value 50.
idum	seed for pseudorandom number sequence.

Details

The Bayesian genomic control statistics with the following parameters,

n	number of loci under consideration
lambdahat	median(of the n trend statistics)/0.46
	Prior for noncentrality parameter Ai is
	Normal(sqrt(lambdahat)kappa,lambdahat*tau2)
kappa	multiplier in prior above, set at 1.6 * sqrt(log(n))
tau2	multiplier in prior above
epsilon	prior probability a marker is associated, set at 10/n
ngib	number of cycles for the Gibbs sampler after burn in
burn	number of cycles for the Gibbs sampler to burn in

Armitage's trend test along with the posterior probability that each marker is associated with the disorder is given. The latter is not a p-value but any value greater than 0.5 (pout) suggests association.

Value

The returned value is a list containing:

• deltot the probability of being an outlier.

• x2 the \chi^2 statistic.

• A the A vector.

Note

Adapted from gcontrol by Bobby Jones and Kathryn Roeder, use -Dexecutable for standalone program, function getnum in the original code needs \

Author(s)

Bobby Jones, Jing Hua Zhao

Source

https://www.cmu.edu/dietrich/statistics-datascience/index.html

References

\insertRef

devlin99gap

Examples

## Not run: 
test<-c(1,2,3,4,5,6,  1,2,1,23,1,2, 100,1,2,12,1,1, 
        1,2,3,4,5,61, 1,2,11,23,1,2, 10,11,2,12,1,11)
test<-matrix(test,nrow=6,byrow=T)
gcontrol(test)

## End(Not run)

gap documentation built on Aug. 26, 2023, 5:07 p.m.