findLinearCombos: Determine linear combinations in a matrix

In caret: Classification and Regression Training

Description

Enumerate and resolve the linear combinations in a numeric matrix

Usage

findLinearCombos(x)

Arguments

x	a numeric matrix

Details

The QR decomposition is used to determine if the matrix is full rank and then identify the sets of columns that are involved in the dependencies.

To "resolve" them, columns are iteratively removed and the matrix rank is rechecked.

The trim.matrix function in the subselect package can also be used to accomplish the same goal.

Value

a list with elements:

linearCombos	If there are linear combinations, this will be a list with elements for each dependency that contains vectors of column numbers.
remove	a list of column numbers that can be removed to counter the linear combinations

Author(s)

Kirk Mettler and Jed Wing (enumLC) and Max Kuhn (findLinearCombos)

See Also

trim.matrix

Examples

testData1 <- matrix(0, nrow=20, ncol=8)
testData1[,1] <- 1
testData1[,2] <- round(rnorm(20), 1)
testData1[,3] <- round(rnorm(20), 1)
testData1[,4] <- round(rnorm(20), 1)
testData1[,5] <- 0.5 * testData1[,2] - 0.25 * testData1[,3] - 0.25 * testData1[,4]
testData1[1:4,6] <- 1
testData1[5:10,7] <- 1
testData1[11:20,8] <- 1

findLinearCombos(testData1)

testData2 <- matrix(0, nrow=6, ncol=6)
testData2[,1] <- c(1, 1, 1, 1, 1, 1)
testData2[,2] <- c(1, 1, 1, 0, 0, 0)
testData2[,3] <- c(0, 0, 0, 1, 1, 1)
testData2[,4] <- c(1, 0, 0, 1, 0, 0)
testData2[,5] <- c(0, 1, 0, 0, 1, 0)
testData2[,6] <- c(0, 0, 1, 0, 0, 1)

findLinearCombos(testData2)

Example output

Loading required package: lattice
Loading required package: ggplot2
$linearCombos
$linearCombos[[1]]
[1] 5 2 3 4

$linearCombos[[2]]
[1] 8 1 6 7


$remove
[1] 5 8

$linearCombos
$linearCombos[[1]]
[1] 3 1 2

$linearCombos[[2]]
[1] 6 1 4 5


$remove
[1] 3 6

caret documentation built on May 2, 2019, 5:47 p.m.