vimp_ci: Confidence intervals for variable importance

In bdwilliamson/npvi: Perform Inference on Algorithm-Agnostic Variable Importance

Confidence intervals for variable importance

Description

Compute confidence intervals for the true variable importance parameter.

Usage

vimp_ci(est, se, scale = "identity", level = 0.95, truncate = TRUE)

Arguments

est	estimate of variable importance, e.g., from a call to vimp_point_est.
se	estimate of the standard error of est, e.g., from a call to vimp_se.
scale	scale to compute interval estimate on (defaults to "identity": compute Wald-type CI).
level	confidence interval type (defaults to 0.95).
truncate	truncate CIs to have lower limit at (or above) zero?

Details

See the paper by Williamson, Gilbert, Simon, and Carone for more details on the mathematics behind this function and the definition of the parameter of interest.

Value

The Wald-based confidence interval for the true importance of the given group of left-out covariates.

bdwilliamson/npvi documentation built on Feb. 1, 2024, 10:46 p.m.