BoostVimp.plot: Variable and Interaction Importance Plots for a LBoost Model

Description

Dot chart of variable and/or interaction importance for the variables/interactions with the largest magnitude variable importance scores.

Usage

BoostVimp.plot(fit, num=10, pred, norm=TRUE, titles=TRUE)

Arguments

fit

an object of class LBoost.

num

number of variables/interactions to be included on plot.

pred

logical. If TRUE, a variable importance plot is constructed for individual variables.

norm

logical. If TRUE, variable/interaction importance scores are normalized such that the largest importance score takes value one and all other values are scaled accordingly.

titles

logical. If FALSE, titles are not included on the plot.

Value

pred=TRUE is specified, one plot will be of the largest magnitude individual variable importance scores. Note, pred.imp must also have been specified as TRUE when running LBoost to be able to generate this plot. A plot for each type of interaction importance measure will also be generated if PI.imp="Both" when running LBoost. If only "Permutation" or "AddRemove" was specified for PI.imp, one plot will be generated for the interaction importance type specified in LBoost.

References

Wolf, B.J., Slate, E.H., Hill, E.G. (2010) Logic Forest: An ensemble classifier for discovering logical combinations of binary markers. Bioinformatics.

Examples

Run this code

data(LBoost.fit)

#Plot of top 10 predictors based on variable importance from the LBoost
#model LBoost.fit
BoostVimp.plot(fit=LBoost.fit, num=10, pred=TRUE, norm=TRUE, titles=TRUE)