subMppca: subMppca

In VBmix: Variational Bayesian Mixture Models

Description

removes unused components and factor columns from model.

Usage

subMppca(model, prune = FALSE, thres = 2.001, quick = FALSE, noxmean = TRUE)

Arguments

model	MPPCA model to be shrinked.
prune	if TRUE, void factor columns are removed.
thres	threshold for component selection. A components is selected iif alpha > thres.
quick	influences method for void factor columns detection. If FALSE, a KL-based criterion is employed (more accurate). If TRUE, column norms are used (useful for very high dimensional data sets).
noxmean	should always be set to TRUE.

Value

shrinked MPPCA model.

Author(s)

Pierrick Bruneau

See Also

mppca newMppca

Examples

# use a subsample of pendat, for runtime (packaging) needs.
temp <- mppca(pendat[sample(1:2000,150),], 15, qmax=8, maxit=20)
temp2 <- subMppca(temp, prune=TRUE, quick=TRUE)

VBmix documentation built on May 30, 2017, 2:34 a.m.