Autoplot Method for LOCO-MP Feature Importance — autoplot.loco_mp

Generates a ggplot2 forest/lollipop chart displaying Leave-One-Covariate-Out importance scores, confidence intervals, and significance thresholds.

Usage

# S3 method for class 'loco_mp_coxnet'
autoplot(object, ...)

Arguments

object: A loco_mp_coxnet object.
...: Additional arguments.

Value

A ggplot2 plot object.

Details

The bars are the two-sided 1 - alpha confidence intervals, which are not adjusted for multiple testing, but a predictor is coloured "Significant" when its adjusted one-sided p-value is below alpha, not when its bar excludes 0. The two criteria can disagree in either direction: with p_adjust = "none", a "Significant" predictor whose p-value is between alpha / 2 and alpha has a bar that still crosses 0, and after a correction such as Bonferroni a bar that excludes 0 can be "Not Significant".