createData: Generates data as in the paper

In dajmcdon/cplr: Compressed and penalized linear regression

Description

Generates data as in the paper

Usage

createData(ntrain, ntest, p, covType, rho, bmethod, SNR, noisesd, ...)

Arguments

ntrain	number of training samples
ntest	number of testing samples
p	number of predictors
covType	for use with generateCov
rho	for use with generateCov
bmethod	for use with generateb
SNR	if not na, sets the standard deviation of the noise to √{b'Cov b} / SNR
noisesd	the standard deviation of the noise if SNR=NA
...	optional arguments passed to generateb

Value

A list with components Xtrain, Ytrain, Xtest, Ytest, bstar, and sig giving respectively the training data, testing data, population regression coefficients, and noise standard deviation. The matrix Xtest is the same as the training data adjusted to the correct size (either by truncating or recycling) so that predictions are at the same design points. Note that this function is simply a wrapper of the various generate* functions for ease of simulation. If more flexibility is desired, use those directly or similar.

Examples

createData(2500, 2, 250, 'corrUniform', .8, 'constant', NA, 2500-250-1)

dajmcdon/cplr documentation built on May 14, 2019, 3:29 p.m.