rjd3highfreq provides functions for seasonal adjustment of +high-frequency data displaying multiple, non integer periodicities. +Pre-adjustment with extended airline model and Arima Model Based +decomposition.
+Running rjd3 packages requires Java 17 or higher. +How to set up such a configuration in R is explained here
+You can install the development version of +rjd3highfreq from GitHub with:
+# Install development version from GitHub
+# install.packages("remotes")
+remotes::install_github("rjdverse/rjd3highfreq")## Import of data
+df_daily <- read.csv2("https://raw.githubusercontent.com/TanguyBarthelemy/Tsace_RJD_Webinar_Dec22/b5fcf6b14ae47393554950547ef4788a0068a0f6/Data/TS_daily_births_franceM_1968_2020.csv")
+
+# Creation of log variables to multiplicative model
+df_daily$log_births <- log(df_daily$births)
+df_daily$date <- as.Date(df_daily$date)Plot of the raw series:
+
+
+Preparation of the calendar with the package +rjd3toolkit:
+# French calendar
+frenchCalendar <- rjd3toolkit::national_calendar(days = list(
+ rjd3toolkit::fixed_day(7, 14), # Bastille Day
+ rjd3toolkit::fixed_day(5, 8, validity = list(start = "1982-05-08")), # End of 2nd WW
+ rjd3toolkit::special_day('NEWYEAR'),
+ rjd3toolkit::special_day('MAYDAY'), # 1st may
+ rjd3toolkit::special_day('EASTERMONDAY'),
+ rjd3toolkit::special_day('ASCENSION'),
+ rjd3toolkit::special_day('WHITMONDAY'),
+ rjd3toolkit::special_day('ASSUMPTION'),
+ rjd3toolkit::special_day('ALLSAINTSDAY'), # Toussaint
+ rjd3toolkit::special_day('ARMISTICE'), # End of 1st WW
+ rjd3toolkit::special_day('CHRISTMAS'))
+)Creation of the calendar regressor in a matrix with the package +rjd3toolkit:
+# Calendar regressor matrix
+cal_reg <- rjd3toolkit::holidays(
+ calendar = frenchCalendar,
+ start = "1968-01-01", length = nrow(df_daily),
+ type = "All", nonworking = 7L)
+
+colnames(cal_reg) <- c("14th_july", "8th_may", "1st_jan", "1st_may",
+ "east_mon", "asc", "pen_mon",
+ "15th_aug", "1st_nov", "11th_nov", "Xmas")Preprocessing with the function
+fractionalAirlineEstimation:
pre_pro <- fractionalAirlineEstimation(
+ y = df_daily$births,
+ x = cal_reg,
+ periods = 7, # weekly frequency
+ outliers = c("ao", "wo"), log = TRUE, y_time = df_daily$date)
+
+print(pre_pro)
+#> Number of observations: 19359
+#> Start: 1968-01-01
+#> End: 2020-12-31
+#>
+#> Estimate MA parameters:
+#> MA_parameter Coef Coef_SE Tstat
+#> Theta(1) 0.7620698 0.005571472 136.7807
+#> Theta(period = 7) 0.9731793 0.001413477 688.5002
+#>
+#> Number of calendar regressors: 11 , Number of outliers : 7
+#>
+#> TD regressors coefficients:
+#> Variable Coef Coef_SE Tstat
+#> 14th_july -0.1226 0.0047 -26.0615
+#> 8th_may -0.1419 0.0054 -26.3419
+#> 1st_jan -0.2223 0.0047 -47.3511
+#> 1st_may -0.1225 0.0047 -26.2643
+#> east_mon -0.1891 0.0046 -40.7635
+#> asc -0.1726 0.0046 -37.1949
+#> pen_mon -0.1900 0.0046 -40.9429
+#> 15th_aug -0.1181 0.0047 -25.3461
+#> 1st_nov -0.1503 0.0046 -32.5662
+#> 11th_nov -0.1238 0.0046 -26.8142
+#> Xmas -0.2310 0.0046 -49.7435
+#>
+#> Outliers coefficients:
+#> Variable Coef Coef_SE Tstat
+#> WO.1999-12-31 -0.1762 0.0226 -7.7916
+#> AO.1995-08-15 -0.2224 0.0340 -6.5503
+#> WO.1999-12-24 -0.1447 0.0226 -6.3981
+#> AO.2012-01-01 0.2098 0.0340 6.1786
+#> AO.1998-07-14 -0.2101 0.0340 -6.1880
+#> AO.1997-07-14 -0.2092 0.0340 -6.1602
+#> AO.1995-05-01 -0.2042 0.0340 -6.0146
+#>
+#> Sum of square residuals: 25.17 on 19330 degrees of freedom
+#> Log likelihood = 3.682e+04,
+#> aic = -7.361e+04,
+#> aicc = -7.361e+04,
+#> bic(corrected for length) = -6.635
+#> Hannan–Quinn information criterion = -7.355e+04
+
+plot(x = pre_pro,
+ from = as.Date("2000-01-01"), to = as.Date("2000-12-31"),
+ main = "French births in 2000")
+
+Decomposition with the AMB (Arima Model Based) algorithm:
+# Decomposition with weekly pattern
+amb.dow <- rjd3highfreq::fractionalAirlineDecomposition(
+ y = pre_pro$model$linearized, # linearized series from preprocessing
+ period = 7,
+ log = TRUE, y_time = df_daily$date)
+
+# Extract day-of-year pattern from day-of-week-adjusted linearised data
+amb.doy <- rjd3highfreq::fractionalAirlineDecomposition(
+ y = amb.dow$decomposition$sa, # DOW-adjusted linearised data
+ period = 365.2425, # day of year pattern
+ log = TRUE, y_time = df_daily$date)Plot:
+ +
plot(amb.dow, main = "Weekly pattern - January 2018",
+ from = as.Date("2018-01-01"),
+ to = as.Date("2018-01-31"))
plot(amb.doy, main = "Weekly pattern - 2000 - 2002",
+ from = as.Date("2000-01-01"),
+ to = as.Date("2002-12-31"))
Perform an Arima Model Based (AMB) decomposition on several +periodcities at once:
+amb.multi <- rjd3highfreq::multiAirlineDecomposition(
+ y = pre_pro$model$linearized, # input time series
+ periods = c(7, 365.2425), # 2 frequency
+ log = TRUE, y_time = df_daily$date)Plot the comparison between the two AMB methods for the annual +periodicity:
+ +
With the package rjd3x11plus, +you can perform an X-11 like decomposition with any (non integer) +periodicity.
+Any contribution is welcome and should be done through pull requests +and/or issues. pull requests should include updated +tests and updated documentation. If +functionality is changed, docstrings should be added or updated.
+The code of this project is licensed under the European +Union Public Licence (EUPL).
+ + + diff --git a/README.md b/README.md index e9192b3..f09bb49 100644 --- a/README.md +++ b/README.md @@ -23,7 +23,7 @@ You can install the development version of **rjd3highfreq** from ``` r # Install development version from GitHub # install.packages("remotes") -remotes::install_github("rjdemetra/rjd3highfreq") +remotes::install_github("rjdverse/rjd3highfreq") ``` ## Demonstration with the daily french births @@ -220,7 +220,7 @@ plot(amb.multi, main = "2012",
With the package
-[**rjd3x11plus**](https://github.com/rjdemetra/rjd3x11plus), you can
+[**rjd3x11plus**](https://github.com/rjdverse/rjd3x11plus), you can
perform an X-11 like decomposition with any (non integer) periodicity.
## Package Maintenance and contributing
diff --git a/_pkgdown.yml b/_pkgdown.yml
index 70a7bc2..ac6f981 100644
--- a/_pkgdown.yml
+++ b/_pkgdown.yml
@@ -1,4 +1,4 @@
-url: https://rjdemetra.github.io/rjd3highfreq/
+url: https://rjdverse.github.io/rjd3highfreq/
template:
bootstrap: 5