hierarchicalFWER: Hierarchical testing with FWER control

In MLGL: Multi-Layer Group-Lasso

Hierarchical testing with FWER control

Description

Apply hierarchical test for each hierarchy, and test external variables for FWER control at level alpha

Usage

hierarchicalFWER(
  X,
  y,
  group,
  var,
  test = partialFtest,
  Shaffer = FALSE,
  addRoot = FALSE
)

Arguments

X	original data
y	associated response
group	vector with index of groups. group[i] contains the index of the group of the variable var[i].
var	vector with the variables contained in each group. group[i] contains the index of the group of the variable var[i].
test	function for testing the nullity of a group of coefficients in linear regression. The function has 3 arguments: X, the design matrix, y, response, and varToTest, a vector containing the indices of the variables to test. The function returns a p-value
Shaffer	boolean, if TRUE, a Shaffer correction is performed
addRoot	If TRUE, add a common root containing all the groups

Details

Version of the hierarchical testing procedure of Meinshausen for MLGL output. You can use th selFWER function to select groups at a desired level alpha

Value

a list containing:

pvalues

pvalues of the different test (without correction)

adjPvalues

adjusted pvalues

groupId

Index of the group

hierMatrix

Matrix describing the hierarchical tree.

References

Meinshausen, Nicolai. "Hierarchical Testing of Variable Importance." Biometrika 95.2 (2008): 265-78.

See Also

selFWER, hierarchicalFDR

Examples

set.seed(42)
X <- simuBlockGaussian(50, 12, 5, 0.7)
y <- X[, c(2, 7, 12)] %*% c(2, 2, -2) + rnorm(50, 0, 0.5)
res <- MLGL(X, y)
test <- hierarchicalFWER(X, y, res$group[[20]], res$var[[20]])

MLGL documentation built on March 31, 2023, 9:32 p.m.