miceadds (version 1.5-0)

pca.covridge: Principal Component Analysis with Ridge Regularization

Description

Performs a principal component analysis for a dataset while a ridge parameter is added on the diagonal of the covariance matrix.

Usage

pca.covridge(x, ridge = 10^(-10))

Arguments

x

A numeric matrix

ridge

Ridge regularization parameter for the covariance matrix

Value

• A list with following entries:
• loadings
  Matrix of factor loadings
• scores
  Matrix of principal component scores
• sdev
  Vector of standard deviations of factors (square root of eigenvalues)

See Also

Principal component analysis in stats: stats::princomp For calculating first eigenvalues of a symmetric matrix see also sirt::eigenvalues.sirt in the sirt package.

Examples

#############################################################################
# EXAMPLE 1: PCA on imputed internet data
#############################################################################

library(mice)
data(data.internet)
dat <- as.matrix( data.internet)

# single imputation in mice
imp <- mice( dat , m=1 , maxit=10 )

# apply PCA
pca.imp <- pca.covridge( complete(imp) )
  ##   > pca.imp$sdev
  ##      Comp.1    Comp.2    Comp.3    Comp.4    Comp.5    Comp.6    Comp.7 
  ##   3.0370905 2.3950176 2.2106816 2.0661971 1.8252900 1.7009921 1.6379599 

# compare results with princomp
pca2.imp <- princomp( complete(imp) )
  ##   > pca2.imp
  ##   Call:
  ##   princomp(x = complete(imp))
  ##   
  ##   Standard deviations:
  ##      Comp.1    Comp.2    Comp.3    Comp.4    Comp.5    Comp.6    Comp.7 
  ##   3.0316816 2.3907523 2.2067445 2.0625173 1.8220392 1.6979627 1.6350428