regrlin: Linear regression with PRESS leave one out estimate

In gbonte/gbcode: Code from the handbook "Statistical foundations of machine learning"

Linear regression with PRESS leave one out estimate

Description

Linear regression

Usage

regrlin(X, Y, X.ts = NULL, lambda = 0.001)

Arguments

X:	training input
Y:	training output
X.ts:	test input
lambda:	regularization parameter

Value

a list with fields:

• e: training error,

• beta.hat: coeffcients,

• MSE.emp: training MSE,

• sdse.emp: standard deviation of squared error,

• var.hat: estimated noise variance;

• MSE.loo: PRESS MSE,

• sdse.loo: standard deviation of squared leave-one-out error error,

• Y.hat: predicted training output,

• Y.hat.ts: predicted test output,

• e.loo: leave-one-out error

Author(s)

Gianluca Bontempi [e-mail]

References

mlg.ulb.ac.be

Examples

N<-100
n<-10
X<-array(rnorm(N*n),c(N,n))
Y<-sin(X[,1]*X[,2])
R<-regrlin(X,Y,X)

gbonte/gbcode documentation built on Aug. 30, 2024, 1:11 a.m.