camt.fwer: Perform the covariate-adaptive family-wise error rate control

In jchen1981/CAMT: Covariate Adaptive Multiple Testing

Description

The function implements a scalable, flexible, robust and powerful FWER control method for large-scale multiple testing exploiting the auxiliary covariates. It allows both the prior null probability to depend on covariates.

Usage

camt.fwer(
  pvals,
  pi0.var,
  data = data,
  pvals.cutoff = 1e-15,
  alpha = 0.1,
  k.fix = NULL,
  EM.paras = list(iterlim = 50, tol = 1e-05, k.init = NULL, pi0.init = NULL, nlm.iter =
    5),
  trace = FALSE,
  return.model.matrix = FALSE
)

Arguments

pvals	a numeric vector of p-values.
pi0.var	a formula, a vector, a data frame, or a matrix of covariate values for the prior null probability.
data	a data frame containing the covariates, only used when pi0.var are classes of 'formula'.
pvals.cutoff	a numeric value to replace p-values below that value, which is used to increase the stability of the algorithm.
alpha	a numeric value for the significance level. Default is 0.1.
k.fix	a numeric value between 0 and 1. It is the k parameter of the alternative (beta) distribution. If it is null, it will be estimated via EM.
EM.paras	a list of control arguments for the EM algorithm iterliman integer value indicating the maximum number of iterations. tola numeric value giving the tolerance in the relative change in the log likelihood below which the algorithm is considered to be converged. pi0.init, k.inittwo scalars giving the initial guess of the pi0 and k parameter. nlm.iteran integer indicating the allowed maximum iteration in running 'nlm', used to speed up computation.
trace	a logical value indicating whether to print the process.
return.model.matrix	a logical value indicating whether to return the model matrix. Consider setting to FALSE if it's huge.

Value

A list with the elements

call	the call made.
pi0	a vector of the estimated null probabilities.
k	the (estimated) value for the k parameter of the alternative distribution.
pi0.coef	a vector of the coefficients for pi0.
gamma	the estimated gamma value to dichotomize the p-values
loglik	log likelihood.
EM.paras	actually used parameters in EM algorithm.
EM.iter	the number of iteration actually used.
rejection	a vector containing the indices of the rejected hypotheses.

Author(s)

Huijuan Zhou

References

Huijuan Zhou, Xianyang Zhang, Jun Chen. Covariate Adaptive Family-wise Error Control with Applications to Genome-wide Association Studies. Submitted.

Examples

data <- simulate.data(feature.no = 10000, covariate.strength = 'Moderate', covariate.model = 'pi0',
	sig.density = 'Medium', sig.strength = 'L4', cor.struct = 'None')
camt.obj.fwer <- camt.fwer(pvals = data$pvals, pi0.var = data$pi0.var)
plot.camt.fwer(camt.obj.fwer, covariate = as.vector(rank(data$pi0.var)), covariate.name = 'Covariate rank',
	log = TRUE, file.name = 'CovariateModerate.pdf')

jchen1981/CAMT documentation built on Jan. 2, 2021, 1:44 p.m.