register_smacof: Register Seriation Methods from Package smacof

In seriation: Infrastructure for Ordering Objects Using Seriation

Register Seriation Methods from Package smacof

Description

Registers the "MDS_smacof" method for seriate() based on multidemensional scaling using stress majorization and the corresponding "smacof_stress0" criterion implemented in package smacof (de Leeuw & Mair, 2009).

Usage

register_smacof()

Details

Seriation method "smacof" implements stress majorization with several transformation functions. These functions are passed on as the type control parameter. We default to "ratio", which together with "interval" performs metric MDS. "ordinal" can be used for non-metric MDS. See smacof::smacofSym() for details on the control parameters.

The corresponding criterion calles "smacof_stress0" is also registered. There additional parameter type is used to specify the used transformation function. It should agree with the function used for seriation. See smacof::stress0() for details on the stress calculation.

Note: Package smacof needs to be installed.

Value

Nothing.

References

Jan de Leeuw, Patrick Mair (2009). Multidimensional Scaling Using Majorization: SMACOF in R. Journal of Statistical Software, 31(3), 1-30. \Sexpr[results=rd]{tools:::Rd_expr_doi("10.18637/jss.v031.i03")}

See Also

Other seriation: register_DendSer(), register_GA(), register_optics(), register_tsne(), register_umap(), registry_for_seriaiton_methods, seriate_best(), seriate()

Examples

## Not run: 
register_smacof()

get_seriation_method("dist", "MDS_smacof")

d <- dist(random.robinson(20, pre = TRUE))

## use Banded AR form with default clustering (complete-link)
o <- seriate(d, "MDS_smacof", verbose = TRUE)
pimage(d, o)

# recalculate stress for the order
MDS_stress(d, o)

# ordinal MDS. stress needs to be calculated using the correct type with stress0
o <- seriate(d, "MDS_smacof", type = "ordinal", verbose = TRUE)
criterion(d, o, method = "smacof_stress0", type = "ordinal")

## End(Not run)

seriation documentation built on Nov. 27, 2023, 1:07 a.m.