wpstCLASS: Predict values using new time series values via a...

In wavethresh: Wavelets Statistics and Transforms

Predict values using new time series values via a non-decimated wavelet packet discrimination object.

Description

Given a timeseries (timeseries) and another time series of categorical values (groups) the makewpstDO produces a model that permits discrimination of the groups series using a discriminant analysis based on a restricted set of non-decimated wavelet packet coefficients of timeseries. The current function enables new timeseries data, to be used in conjunction with the model to generate new, predicted, values of the groups time series.

Usage

wpstCLASS(newTS, wpstDO)

Arguments

newTS	A new segment of time series values, of the same time series that was used as the dependent variable used to construct the wpstDO object
wpstDO	An object that uses values of a dependent time series to build a discriminatory model of a groups time series. Output from the makewpstDO function

Details

This function performs the same nondecimated wavelet packet (NDWPT) transform of the newTS data that was used to analyse the original timeseries and the details of this transform are stored within the wpstDO object. Then, using information that was recorded in wpstDO the packets with the same level/index are extracted from the new NDWPT and formed into a matrix. Then the linear discriminant variables, again stored in wpstDO are used to form predictors of the original groups time series, ie new values of groups that correspond to the new values of timeseries.

Value

The prediction using the usual R predict.lda function. The predicted values are stored in the class component of that list.

Author(s)

G P Nason

See Also

makewpstDO

Examples

#
# See example at the end of help page for makewpstDO
#

wavethresh documentation built on Nov. 16, 2022, 5:16 p.m.