gdpc

This package provides functions for computing the Generalized Dynamic Principal Components proposed in Peña and Yohai (2016). It also provides a function to automatically choose the number of components and lags, using the criteria described in Peña, Smucler and Yohai (2020).

---

Installation

You can install the stable version from R CRAN.

install.packages('gdpc', dependencies = TRUE)

You can install the development version from GitHub

library(devtools)
devtools::install_github("esmucler/gdpc")

Usage

library(gdpc)
# An example using an artificial data set
T <- 200 
m <- 500
set.seed(1234)
f <- rnorm(T + 1)
x <- matrix(0, T, m)
u <- matrix(rnorm(T * m), T, m)
for (i in 1:m) {
    x[, i] <- 10 * sin(2 * pi * (i/m)) * f[1:T] + 10 * cos(2 * pi * (i/m)) * f[2:(T + 1)] + u[, i]
}
# Compute first DPC using one lag
fit <- gdpc(x, k = 1)
fit
# Plot loadings
par(mfrow = c(1, 2))
plot(fit, which = 'Loadings', which_load = 0, xlab = '', ylab = '') 
plot(fit, which = 'Loadings', which_load = 1, xlab = '', ylab = '') 

# An example using a real data set
# Load and plot Industrial Production Index Data
data(ipi91)
par(mfrow = c(1, 1))
plot(ipi91, plot.type = 'multiple', main = 'Industrial Production Index')
# Choose number of lags among 0,..., 9 using the (default) Leave One Out criterion
# This might take a minute.
gdpc_ipi <- auto.gdpc(ipi91, k_max = 9, niter_max = 1500, ncores = 4)
gdpc_ipi
# Plot the component
plot(gdpc_ipi, which_comp = 1, ylab = '')
# Get reconstruction of the time series and plot
recons <- fitted(gdpc_ipi, 1)
colnames(recons) <- colnames(ipi91)
plot(recons, main = 'Fitted values')

License

This package is free and open source software, licensed under GPL (>= 2).