Skip to contents

Nonmetric multidimensional scaling

Source: R/scaling.R
tda_mdsn.Rd

TDA's mdsn1 (Kruskal) on a distance matrix.

Usage

tda_mdsn(
  d,
  ties = c("ignore", "primary", "secondary"),
  max_iterations = 100,
  restarts = 5,
  ...
)

Arguments

d

symmetric distance matrix.

ties

tie handling. Only "ignore" is implemented in mdsn1 (its primary and secondary projections are commented out in the C, and the option would silently do nothing); the other two values error and point to tda_mds, whose mdsn has them.

max_iterations

iteration cap.

restarts

random starting configurations, TDA's ns=; the one with the lowest stress is returned. Nonmetric MDS on few points has degenerate local minima (clusters of near-coincident points with small stress), and restarts are the guard against them.

...

passed to tda_run.

Value

with direct exports enabled, a list with stress, configuration (n x 2) and output; otherwise the printed output lines.

Examples

d <- as.matrix(dist(cbind(c(0, 0, 3, 3), c(0, 2, 0, 2))))
tda_mdsn(d)
#> <tda_mds_nonmetric>
#> stress: 0 
#> configuration: 4 x 2 matrix
#> (full TDA output in $output, 408 lines)