Guttman-transform majorisation of raw stress from the Torgerson start
with ratio, interval or ordinal (primary ties) optimal scaling and
disparities normalised to \(\sum w\hat d^2 = n(n-1)/2\), stopping when
the stress decrease falls below eps, exactly as
smacof::smacofSym (De Leeuw and Mair 2009). NA dissimilarities or
zero weights are missing. Identical to the Python arm
morie.fn.mdsops.smacof.
Usage
SmacofMds(
delta,
ndim = 2,
type = "ratio",
weights = NULL,
init = NULL,
itmax = 1000L,
eps = 1e-06
)
Arguments
- delta
Symmetric dissimilarity matrix.
- ndim
Dimensions.
- type
"ratio", "interval" or "ordinal".
- weights
Weight matrix.
- init
Starting configuration (default Torgerson).
- itmax
Maximum iterations.
- eps
Convergence tolerance.
Value
List with conf, stress, dhat, confdist, spp, niter.
References
De Leeuw, J. and Mair, P. (2009). Multidimensional scaling
using majorization: SMACOF in R. Journal of Statistical Software 31(3),
1-30.
Examples
D <- rbind(c(0, 3, 4, 6), c(3, 0, 5, 4), c(4, 5, 0, 3), c(6, 4, 3, 0))
SmacofMds(D)$stress
#> [1] 0.04961223
