evalpred: Evaluate prediction performance

In jpvert/kmr: Kernel multitask regression

Description

Computes performance scores of predicted against true responses.

Usage

evalpred(ypred, ytest, type.measure = c("ci", "mse", "cor"))

Arguments

ypred	Matrix or list of matrices (e.g. each corresponding to a tested lambda) of predicted reponses, of dimension nobs x ntask.
ytest	Matrix of true reponses, of dimension nobs x ntask.
type.measure	Measure type for evaluating performance. Possible options are "ci" (concordance index), "mse" (mean squared error), "cor" (pearson correlation). Default is "ci".

Value

A matrix of mean performance scores (averaged over observations), of dimension ntask x nlambda where nlambda denotes the length of the list ypred.

Note

ypred and ytest are not interchangably equivalent for type.measure="ci", as ytest is treated as true responses and used to determine denominator in computing C-index.

jpvert/kmr documentation built on May 20, 2019, 7:56 a.m.