Dynamic Principal Components for Periodically Correlated Functional Time Series

Method extends multivariate and functional dynamic principal components to periodically correlated multivariate time series. This package allows you to compute true dynamic principal components in the presence of periodicity. We follow implementation guidelines as described in Kidzinski, Kokoszka and Jouzdani (2017), in Principal component analysis of periodically correlated functional time series .


Implementation of "Dynamic principal components of periodically correlated functional time series".

Two examples in demo directory:

  • pm10 data from Graz (comparizon with DFPCA paper)
  • simplation with parametrized periodicity

Installation

library("devtools")
install_github("kidzik/pcdpca")

Running a demo

library("pcdpca")
demo("simulation")
demo("pcdpca.pm10")

Usage

Let X be a multivariate time series, a matrix with n observations and d covariates, periodic with period = 2. Then

FF = pcdpca(X, period=2)  # finds the optimal filter
Yhat = pcdpca.scores(X, FF)  # applies the filter
Yhat[,-1] = 0 # forces the use of only one component
Xhat = pcdpca.inverse(Yhat, FF)  # deconvolution
cat(sum((X-Xhat)^2) / sum(X^2)) # variance explained

News

Reference manual

It appears you don't have a PDF plugin for this browser. You can click here to download the reference manual.

install.packages("pcdpca")

0.4 by Lukasz Kidzinski, 4 years ago


Browse source code at https://github.com/cran/pcdpca


Authors: Lukasz Kidzinski [aut, cre] , Neda Jouzdani [aut] , Piotr Kokoszka [aut]


Documentation:   PDF Manual  


Task views: Time Series Analysis, Functional Data Analysis


GPL-3 license


Imports freqdom, fda


See at CRAN