Selection from hierarchical testing with FDR control

Share:

Description

Select groups from hierarchical testing procedure with FDR control (hierarchicalFDR)

Usage

1
selFDR(out, alpha = 0.05, global = TRUE, outer = TRUE)

Arguments

out

output of hierarchicalFDR function

alpha

control level for test

global

if FALSE the provided alpha is the desired level control for each family.

outer

if TRUE, the FDR is controlled only on outer node (rejected groups without rejected children) . If FALSE, it is controlled on the full tree.

Details

See the reference for mode details about the method.

If each family is controlled at a level alpha, we have the following control : FDR control of full tree : alpha * delta * 2 (delta = 1.44) FDR control of outer node : alpha * L * delta * 2 (delta = 1.44)

Value

a list containing :

toSel

vector of boolean. TRUE if the group is selected

groupId

Names of groups

local.alpha

control level for each family of hypothesis

global.alpha

control level for the tree (full tree or outer node)

References

Yekutieli, Daniel. "Hierarchical False Discovery Rate-Controlling Methodology." Journal of the American Statistical Association 103.481 (2008): 309-16.

See Also

hierarchicalFDR

Examples

1
2
3
4
5
6
set.seed(42)
X = simuBlockGaussian(50,12,5,0.7)
y = drop(X[,c(2,7,12)]%*%c(2,2,-2)+rnorm(50,0,0.5))
res = MLGL(X,y)
test = hierarchicalFDR(X, y, res$group[[20]], res$var[[20]])
sel = selFDR (test, alpha = 0.05)