groupfsInf {selectiveInference}

Compute selective p-values for a model fitted by groupfs.

Description

Computes p-values for each group of variables in a model fitted by groupfs. These p-values adjust for selection by truncating the usual \chi^2 statistics to the regions implied by the model selection event. If the sigma to groupfs was NULL then groupfsInf uses truncated F statistics instead of truncated \chi. The sigma argument to groupfsInf allows users to override and use \chi, but this is not recommended unless \sigma can be estimated well (i.e. n > p).

Usage

groupfsInf(obj, sigma = NULL, verbose = TRUE)

Arguments

obj

Object returned by groupfs function

sigma

Estimate of error standard deviation. Default is NULL and in this case groupfsInf uses the value of sigma specified to groupfs.

verbose

Print out progress along the way? Default is TRUE.

Value

An object of class "groupfsInf" containing selective p-values for the fitted model obj. For comparison with fsInf, note that the option type = "active" is not available.

vars

Labels of the active groups in the order they were included.

pv

Selective p-values computed from appropriate truncated distributions.

sigma

Estimate of error variance used in computing p-values.

TC or TF

Observed value of truncated \chi or F.

df

Rank of group of variables when it was added to the model.

support

List of intervals defining the truncation region of the corresponding statistic.