penaltyPolyMARS function

Choice of the penalty parameter for a PolyMARS model

This function fits a PolyMARS model for different values of the penalty parameter and compute criteria.

penaltyPolyMARS(X,Y,test=NULL,graphic=FALSE,K=10,
                Penalty=seq(0,5,by=0.2))

Arguments

• X: a data.frame containing the design of experiments
• Y: a vector containing the response variable
• test: a data.frame containing the design and the response of a test set when available, the prediction criteria will be computed for the test data (default corresponds to no test set)
• graphic: if TRUE the values of the criteria are represented
• K: the number of folds for cross-validation (by default, K=10)
• Penalty: a vector containing the values of the penalty parameter

Returns

A data frame containing - a: the values of the penalty parameter

• R2: the R2 criterion evaluted on the learning set

• m: the size of the selected model

If a test set is available the last row is - R2test: the R2 criterion evaluated on the test set

If no test set is available, criteria computed by K-corss-validation are provided: - Q2: the Q2 evaluated by cross-validation (by default, K=10)

• RMSE CV: RMSE computed by cross-validation

Note that the penalty parameter could be chosen by minimizing the value of the RMSE by cross-validation.

See Also

modelFit, R2 and crossValidation

Author(s)

D. Dupuy

Examples

data(dataIRSN5D)
X       <- dataIRSN5D[,1:5]
Y       <- dataIRSN5D[,6]
data(testIRSN5D)
library(polspline)
Crit    <- penaltyPolyMARS(X,Y,test=testIRSN5D[,-7],graphic=TRUE)