mvtb.ri: Computes the relative influence of each predictor for each...

In patr1ckm/mvtboost: Tree Boosting for Multivariate Outcomes

Description

The relative influence of a predictor is the reduction in sums of squares attributable to splits on individual predictors. It is often expressed as a percent (sums to 100).

Usage

mvtb.ri(model, n.trees = NULL, relative = "col", sort = FALSE, ...)

Arguments

model	mvtb output model
n.trees	number of trees to use. Defaults to the minimum number of trees by CV, test, or training error for each outcome.
relative	How to scale the multivariate influences. If "col", each column sums to 100. If "tot", the whole matrix sums to 100 (a percent). Otherwise, the raw reductions in SSE are returned.
sort	whether or not results should be (reverse) sorted. Defaults to FALSE.
...	Additional arguments passed to gbm::relative_influence

Value

Matrix of (relative) influences.

patr1ckm/mvtboost documentation built on May 24, 2019, 8:21 p.m.