pcs

From DavinciWiki
Jump to: navigation, search


Description

Performs principal component stretch of the given data

Arguments and Return Values

Arguments: A numerical array and options

Return Value: A numerical array

Usage

Syntax: pcs(obj = VAL [, opt = r/v/s] [, axis = x/y/z] [, scale = VAL] [, niter = INT])

The pcs() function performs a Principal Component Stretch on the input data.

The 'opt' parameter selects the statistical method to be used on the input data for the whole analysis. Various values taken by this parameter have the following meaning:

'v' (default) - chooses Principal Component Analysis based on the covariance matrix

'r' - chooses Principal Component Analysis based on the correlation matrix

's' - chooses Principal Component Analysis based on the "sum of cross products" matrix

The 'axis' parameter specifies the dimension along with the vectors are laid out. Say, the input 'obj' is an [m x n x w] matrix and the 'axis' has been specified as 'z' then:

'w' is the dimension of the a single input vector, and there are 'm x n' of them in the image cube.

The default value for this parameter is 'z'.

The 'scale' parameter specifies the scaling factor, if any, to be applied to the stretched vectors during the stretch phase. If 'scale' is not specified, it is assumed to be 1, i.e. no scaling.

The 'niter' parameter specifies the maximum number of iterations allowed during Tridiagonal QL algorithm. If pcs() fails due to 'no convergence' error, try increasing this number.

Examples

dv> a
5x5x5 array of float, bsq format [500 bytes]

dv> pcs_a = pcs(a,v,x)
5x5x5 array of float, bsq format [500 bytes]

DavinciWiki Mini-Nav Bar

Contents


Contact Developers

  • davinci-dev [AT] mars.asu.edu

All other topics

  • See navigation on the left

Related Functions

Recent Core Changes

Modified On: 03-27-2006

Personal tools