plotSubset {cNORM}

Evaluate information criteria for regression model

Description

Plots the information criterion - either Cp (default) or BIC - against the adjusted R square of the feature selection in the modeling process. Both BIC and Mallow's Cp are measures to avoid over-fitting. Please choose the model that has a high information criterion, while modeling the original data as close as possible. R2 adjusted values of ~ .99 might work well, depending on your scenario. In other words: Look out for the elbow in the curve and choose th model where the information criterion begins to drop. Nonetheless, inspect the according model with plotPercentiles(data, group) to visually inspect the course of the percentiles. In the plot, Mallow's Cp is log transformed and the BIC is always highly negative. The R2 cutoff that was specified in the bestModel function is displayed as a dashed line.

Usage

plotSubset(model, type = 0, index = FALSE)

Arguments

model

The regression model from the bestModel function or a cnorm object

type

Type of chart with 0 = adjusted R2 by number of predictors, 1 = log transformed Mallow's Cp by adjusted R2, 2 = Bayesian Information Criterion (BIC) by adjusted R2, 3 = Root Mean Square Error (RMSE), 4 = Residual Sum of Squares by number, 5 = F-test statistic for consecutive models and 6 = p-value for model tests of predictors

index

add index labels to data points

See Also

bestModel, plotPercentiles, printSubset

Other plot: plot.cnorm(), plotDensity(), plotDerivative(), plotNormCurves(), plotNorm(), plotPercentileSeries(), plotPercentiles(), plotRaw()

Examples

# Compute model with example data and plot information function
cnorm.model <- cnorm(raw = elfe$raw, group = elfe$group)
plotSubset(cnorm.model)