The package augments Mehmet Balcilar’s
mFilter package by
providing a function for the calculation of confidence bands for the
trend component of the two-sided Hodrick-Prescott (HP) filter as
proposed by Giles
(2013).
Note that the construction of such confidence bands is recommended only for stationary times series.
# install.packages("devtools")
devtools::install_github("franzmohr/hpconf")The following code reproduces figures 3 and 4 in Giles (2013) in one graph.
# Load packages
library(alfred)
library(mFilter)
library(hpconf)
# Download data
unemp <- get_fred_series(series_id = "LNS14100000", series_name = "value",
observation_start = "1968-01-01", observation_end = "2012-03-01")
# Transform data into time-series object
unemp <- ts(unemp$value, start = 1968, frequency = 12)
# HP-filter
hp_unemp <- hpfilter(unemp, freq = 14400)
# Obtain confidence bands
hp_unemp_conf <- hpconf(hp_unemp)
# Plot
plot(hp_unemp_conf, main = "US unemployment rate", ylab = "", xlab = "", ylim = c(0, 12))
Balcilar, M. (2018). mFilter: Miscellaneous Time Series Filters. R package version 0.1-4. https://CRAN.R-project.org/package=mFilter
Giles, D. E. (2013). Constructing confidence bands for the Hodrick-Prescott filter. Applied Economics Letters (20)5, 480-484. https://doi.org/10.1080/13504851.2012.714057