Best1DCols: Extract the best (one-dimensional) nondecimated WP packets

Best1DColsR Documentation

Extract the best (one-dimensional) nondecimated WP packets

Description

This function takes the whole set of nondecimated wavelet packets and selects those packets that correlate best with the "response" groups. The idea is to reduce the large dimensionality (number of packets) into something more manageable which can then be fed into a proper discriminator.

Usage

Best1DCols(w2d, mincor=    0.69999999999999996)

Arguments

w2d

An object that gets returned from a call to the wpst2discr function which turns a wpst class object into a regular multivariate matrix

mincor

The threshold above which variables (packets) get included into the final mix if their correlation with the groups variable is higher than this value.

Details

This function is not intended for direct user use. In this function, the w2d object contains a matrix where each column contains the coefficients of a single packet from a non-decimated wavelet packet transform. The number of rows of the matrix is the same as the original time series and hence each column can be correlated with a separate groups variable that contains the group membership of a separate variable which changes over time. Those packet columns that have correlation greater than the mincor value are extracted and returned in the BasisMatrix item of the returned list.

Value

A list with the following components:

nlevelsWT

The number of levels of the nondecimated wavelet packet encapsulator, w2d

BasisMatrix

The highest correlating packets, sorted according to decreasing correlation

level

The levels corresponding to the selected packets

pkt

The packet indices corresponding to the selected packets

basiscoef

The sorted correlations

groups

The groups time series

Author(s)

G P Nason

See Also

makewpstDO,wpst2discr


wavethresh documentation built on Sept. 11, 2024, 9:33 p.m.