testLmRsquared {rotRPackage}R Documentation

$R^2$ test for a linear model.

Description

This ROT function, called from a Test C++ object, is given two samples, a scalar and a parameter vector. It predicts the values corresponding to the explanatory variables through the linear model, then computes the $R^2$. It is tested against the scalar, then the function returns the result of the test and the $R^2$ value.

Usage

testLmRsquared(x, beta, y, testLevel = 0.95)

Arguments

x A m-by-n matrix containing the explanatory variables.
beta A n-by-1 vector containng the linear model parameters.
y A n-by-1 vector containng the response variables.
testLevel the test level. (scalar in [0:1])

Details

As it is not asked in LinearModel.getPredict(), no prediction interval is returned; it is up to the user to be careful about that. It is also to noted that the sample is not assumed to contain the '1's corresponding to the intercept parameter.

Value

A list is returned, containing two scalars ,

testResult A scalar simulating a boolean (easier for Rserve)
valueRSquared A scalar.

Author(s)

Pierre-Matthieu Pair, Régis Lebrun.

Examples

set.seed(1)
x <- matrix(runif(40), 10, 4)
r <- matrix(c(1,2,3,4), 4, 1)
y <- x %*% r + matrix(rnorm(10, 0, 0.05), 10, 1)
LM <- computeLinearModel(x, y)
testLmRsquared(x, LM$parameterEstimate, y)

[Package rotRPackage version 1.4.3 Index]