Pcaengine

From Eigenvector Documentation Wiki

Jump to: navigation, search

Contents

Purpose

Principal components analysis computational engine.

Synopsis

[ssq,datarank,loads,scores,msg] = pcaengine(data,ncomp,options)

Description

This function is intended primarily for use as the engine behind other more full featured PCA programs. The only required input is a data matrix.

Optional inputs include the number of principal components desired in the output ncomp, and a structure containing optional inputs options. If the number of components ncomp is not specified, the routine will return components up to the rank of the data datarank.

The outputs are the variance or sum-of-squares captured table ssq, mathematical rank of the data datarank, principal component loadings loads, principal component scores scores, and a text variable containing any warning messages msg.

To enhance speed, the routine is written so that only the specified outputs are computed.

Inputs

  • data = input data matrix

Optional Inputs

  • ncomp = number of principal components desired in the output
  • options = discussed below.

Outputs

  • ssq = The outputs are the variance or sum-of-squares captured table
  • datarank = mathematical rank of the data
  • loads = principal component loadings
  • scores = principal component scores
  • msg = a text variable containing any warning messages

Options

options = a structure array with the following fields:

  • display: [ 'off' | {'on'} ], governs level of display to command window,
  • algorithm: [ {'regular'} | 'big' | 'auto'], tells which algorithm to use,
  • 'regular', uses an SVD and calculates all eigenvectors and eigenvalues,
  • 'big', calculates the "economy size" SVD, and
  • 'auto', checks the size of the data matrix and automatically chooses between 'regular' and 'big'

The default options can be retreived using: options = pcaengine('options');.

See Also

analysis, estimatefactors, evolvfa, ewfa, explode, parafac, pca, ssqtable, subgroupcl

Views
Personal tools