PCAtools (version 1.3.6)

findElbowPoint: Find the elbow point

Description

Find the elbow point in the curve of variance explained by each successive PC. This can be used to determine the number of PCs to retain.

Usage

findElbowPoint(variance)

Arguments

variance

Numeric vector containing the variance explained by each PC. Should be monotonic decreasing.

Value

An integer scalar specifying the number of PCs at the elbow point.

Details

Our assumption is that each of the top PCs capturing real signal should explain much more variance than the remaining PCs. Thus, there should be a sharp drop in the percentage of variance explained when we move past the last “important” PC. This elbow point lies at the base of this drop in the curve, and is identified as the point that maximizes the distance from the diagonal.

Examples

# NOT RUN {
col <- 20
row <- 1000
mat <- matrix(rexp(col*row, rate = 1), ncol = col)

# Adding some structure to make it more interesting.
mat[1:100,1:3] <- mat[1:100,1:3] + 5
mat[1:100+100,3:6] <- mat[1:100+100,3:6] + 5
mat[1:100+200,7:10] <- mat[1:100+200,7:10] + 5
mat[1:100+300,11:15] <- mat[1:100+300,11:15] + 5

p <- pca(mat)
chosen <- findElbowPoint(p$variance)

plot(p$variance)
abline(v=chosen, col="red")
# }