resemble (version 3.0.1)

diss_pca: PCA dissimilarity method constructor

Description

Creates a configuration object for computing dissimilarity based on Mahalanobis distance in PCA score space.

Usage

diss_pca(
  ncomp = ncomp_by_var(0.01),
  method = c("pca", "pca_nipals"),
  center = TRUE,
  scale = FALSE,
  return_projection = FALSE
)

Value

An object of class c("diss_pca", "diss_method").

Arguments

ncomp

Component selection method. Can be:

• A positive integer for a fixed number of components

• ncomp_fixed(n): explicit fixed selection

• ncomp_by_var(min_var): retain components explaining at least min_var variance each (default: ncomp_by_var(0.01))

• ncomp_by_cumvar(min_cumvar): retain components until cumulative variance reaches min_cumvar

• ncomp_by_opc(): optimize using side information (Yr required in dissimilarity())

method

Character. PCA algorithm: "pca" (default, SVD-based) or "pca_nipals" (NIPALS algorithm).

center

Logical. Center data before projection? Default TRUE.

scale

Logical. Scale data before projection? Default FALSE.

return_projection

Logical. Return the projection object? Default FALSE.

See Also

Component selection: ncomp_by_var, ncomp_by_cumvar, ncomp_by_opc, ncomp_fixed

Other dissimilarity methods: diss_pls, diss_correlation, diss_euclidean, diss_cosine, diss_mahalanobis

Examples

# Fixed number of components
diss_pca(ncomp = 10)
diss_pca(ncomp = ncomp_fixed(10))

# Retain components explaining >= 1% variance each (default)
diss_pca(ncomp = ncomp_by_var(0.01))

# Retain components until 99% cumulative variance
diss_pca(ncomp = ncomp_by_cumvar(0.99))

# Optimize using side information (requires Yr)
diss_pca(ncomp = ncomp_by_opc(40))
diss_pca(ncomp = ncomp_by_opc())

# NIPALS algorithm (useful for very large matrices)
diss_pca(ncomp = 10, method = "pca_nipals")