score: Compute one of several loss metrics on a new data set

In jolars/prague: Generalized Linear Models Regularized with the Sorted L1-Norm

Description

This function is a unified interface to return various types of loss for a model fit with owl().

Usage

score(object, x, y, measure)

## S3 method for class 'OwlGaussian'
score(object, x, y, measure = c("mse", "mae"))

## S3 method for class 'OwlBinomial'
score(object, x, y, measure = c("mse", "mae", "deviance", "misclass", "auc"))

## S3 method for class 'OwlMultinomial'
score(object, x, y, measure = c("mse", "mae", "deviance", "misclass"))

## S3 method for class 'OwlPoisson'
score(object, x, y, measure = c("mse", "mae"))

Arguments

object	an object of class "Owl"
x	feature matrix
y	response
measure	type of target measure. "mse" returns mean squared error. "mae" returns mean absolute error, "misclass" returns misclassification rate, and "auc" returns area under the ROC curve.

Value

The measure along the regularization path depending on the value in measure.

Examples

x <- subset(infert, select = c("induced", "age", "pooled.stratum"))
y <- infert$case

fit <- owl(x, y, family = "binomial")
score(fit, x, y, measure = "auc")

jolars/prague documentation built on March 4, 2020, 7:13 p.m.