Implements a two–segment piece-wise linear method to identify the knee
(a.k.a. elbow) in a curve such as the reconstruction-error profile
obtained when varying the number of archetypes.
Usage
find_knee_pt(y, x = seq_along(y), make_plot = FALSE)
Arguments
- y
Numeric vector of \(y\)-values (must have ≥ 3 points).
- x
Numeric vector of \(x\)-values (defaults to
seq_along(y)). Need not be sorted.
- make_plot
Logical; if TRUE, a ggplot2 graphic is
produced showing the curve and the detected knee.
Value
A list with components
- knee_x
The \(x\)-coordinate of the knee.
- idx
Integer index (1-based) of that knee in the supplied
vectors.
- error_curve
Numeric vector of summed absolute-error values
for every tested breakpoint (length = length(y) - 2).
Examples
ks <- 2:25
err <- exp(-ks) + 0.05 * rnorm(length(ks))
knee_pt(err, ks, make_plot = TRUE)
#> Error in knee_pt(err, ks, make_plot = TRUE): could not find function "knee_pt"