1 - R et JDemetra+

Désaisonnaliser une série temporelle

Auteur

Alain Quartier-la-Tente

L’objectif de ce TP est d’apprendre à manipuler JDemetra+ sous R à travers le package RJDemetra.

Pour manipuler JDemetra+ sous R il y a actuellement deux façons :

  1. Utiliser le JWSACruncher qui permet, à partir de la console, de mettre à jour un workspace JDemetra+ et d’exporter les résultats sans devoir ouvrir le logiciel. Pour faciliter son utilisation depuis R, le package rjwsacruncher peut être utilisé (voir TP5 - JDemetra+ en production).

  2. Utiliser le package RJDemetra qui permet d’effectuer des désaisonnalisations avec les mêmes algorithmes et paramètres que JDemetra+ et de manipuler des workspaces.

Dans ce TP on utilisera les données du package RJDemetra mais n’hésitez pas à utiliser vos propres séries !

Pour faire de la désaisonnalisation sous R il existe plusieurs packages :

RJDemetra est sur le CRAN et se base sur les librairies Java de JDemetra+. Pour l’utiliser il faut avoir Java 8 ou plus. En cas de problème d’installation voir la page : https://github.com/jdemetra/rjdemetra/wiki/Installation-manual.

Le package a aussi un site web (https://jdemetra.github.io/rjdemetra/).

RJDemetra permet :

Une nouvelle version de RJDemetra est en cours de développement autour de la version 3.0 de JDemetra+. Elle nécessite toutefois d’avoir une version de Java supérieure à la version 17 et elle n’est pas sur le CRAN. Les fonctionnalités de RJDemetra (et les nouvelles) sont divisées en plusieurs packages disponibles sous https://github.com/rjdemetra. Dans cette formation, nous utiliserons les packages suivants RJDemetra, rjd3toolkit, rjd3tramoseats, rjd3providers, rjdemetra3 :

packages_to_install <- c("RJDemetra", "remotes")

packages <- packages_to_install[! packages_to_install %in% installed.packages()[,"Package"]]
if (length(packages) > 0) {
    install.packages(packages)
}
packages_to_install_git <- c("rjd3toolkit", "rjd3x13", "rjd3tramoseats", "rjd3providers", "rjdemetra3")
packages_git <- packages_to_install_git[! packages_to_install_git %in% installed.packages()[,"Package"]]

if (length(packages_git) > 0) {
    # # Configurer si besoin le proxy
    # proxy <- "proxy_a_definir"
    # Sys.setenv(HTTPS_PROXY = proxy)
    remotes::install_github(
        sprintf("rjdemetra/%s", packages_git),
        # option utile dans certaines installations portables de Java :
        INSTALL_opts = "--no-multiarch")
}
library(RJDemetra)

1 RJDemetra v2

Commencez par charger le package RJDemetra :

library(RJDemetra)

1.1 Créer une specification

Dans les prochains exercices, la série utilisée sera ipi_c_eu[, "FR"] qui est l’IPI français. Vous pouvez bien sûr adapter le code pour utiliser vos propres séries. Les fonctions utilisées seront x13(), x13_spec(), regarima_x13, regarima_x13_spec ou regarima. Le détail des spécifications pré-définies par JDemetra+ sont disponibles ici.

Exercice

Faire la désaisonnalisation d’une série avec X-13 avec la spécification suivante :

  • détection automatique du schéma de décomposition, des outliers et du modèle ARIMA ;

  • une correction des jours ouvrables “working days” et un effet graduel de Pâques.

Faire ensuite un graphique avec la série brute et la série désaisonnalisée.

Utiliser la spécification RSA4c pour la désaisonnalisation.

Si le modèle créé s’appelle mysa, regarder les valeurs de mysa$final, mysa$final$series et mysa$final$forecasts.

library(RJDemetra)
mysa <- x13(ipi_c_eu[, "FR"], spec = "RSA4c")
mysa
RegARIMA
y = regression model + arima (2, 1, 1, 0, 1, 1)
Log-transformation: no
Coefficients:
          Estimate Std. Error
Phi(1)     0.05291      0.108
Phi(2)     0.18672      0.074
Theta(1)  -0.52137      0.103
BTheta(1) -0.66132      0.042

             Estimate Std. Error
Week days      0.6927      0.031
Leap year      2.0903      0.694
Easter [1]    -2.5476      0.442
TC (4-2020)  -35.6481      2.092
AO (3-2020)  -21.1492      2.122
AO (5-2011)   13.1869      1.810
LS (11-2008)  -9.2744      1.758
LS (1-2009)   -7.2838      1.756


Residual standard error: 2.193 on 346 degrees of freedom
Log likelihood = -795.1, aic =  1616 aicc =  1617, bic(corrected for length) = 1.767



Decomposition
Monitoring and Quality Assessment Statistics:
      M stats
M(1)    0.127
M(2)    0.079
M(3)    1.094
M(4)    0.558
M(5)    1.093
M(6)    0.022
M(7)    0.085
M(8)    0.242
M(9)    0.064
M(10)   0.261
M(11)   0.247
Q       0.355
Q-M2    0.389

Final filters: 
Seasonal filter:  3x5
Trend filter:  13 terms Henderson moving average


Final
Last observed values
             y        sa        t           s           i
Jan 2020 101.0 102.87273 103.0457  -1.8727280  -0.1730003
Feb 2020 100.1 103.69025 103.0626  -3.5902540   0.6276448
Mar 2020  91.8  82.69170 103.2654   9.1083000 -20.5736602
Apr 2020  66.7  66.55184 103.6945   0.1481625 -37.1426277
May 2020  73.7  79.28883 104.1379  -5.5888279 -24.8490764
Jun 2020  98.2  87.35362 104.4539  10.8463757 -17.1002284
Jul 2020  97.4  92.26057 104.5518   5.1394323 -12.2912806
Aug 2020  71.7  97.54392 104.3369 -25.8439193  -6.7929846
Sep 2020 104.7  97.75728 103.8361   6.9427184  -6.0788659
Oct 2020 106.7  97.87016 103.1969   8.8298396  -5.3267823
Nov 2020 101.6 100.01475 102.6601   1.5852540  -2.6453446
Dec 2020  96.6  99.61740 102.4081  -3.0173983  -2.7907314

Forecasts:
               y_f     sa_f      t_f         s_f        i_f
Jan 2021  94.29728 101.0937 102.3876  -6.7963909 -1.2939310
Feb 2021  97.89298 101.6869 102.4525  -3.7939488 -0.7655396
Mar 2021 113.65190 102.1478 102.4593  11.5041248 -0.3115701
Apr 2021 102.34532 102.1806 102.3419   0.1647274 -0.1612727
May 2021  96.14552 101.6436 102.1712  -5.4980759 -0.5276194
Jun 2021 112.15785 101.2156 102.0241  10.9422549 -0.8085520
Jul 2021 104.38495 101.5439 101.9635   2.8410057 -0.4195241
Aug 2021  79.02901 102.3820 102.0412 -23.3530134  0.3408366
Sep 2021 109.40288 102.3705 102.1977   7.0323700  0.1728109
Oct 2021 108.22472 101.8558 102.3655   6.3689511 -0.5096780
Nov 2021 106.22014 102.4265 102.5266   3.7936049 -0.1000754
Dec 2021  99.67523 102.9462 102.6812  -3.2709156  0.2649189


Diagnostics
Relative contribution of the components to the stationary
portion of the variance in the original series,
after the removal of the long term trend
 Trend computed by Hodrick-Prescott filter (cycle length = 8.0 years)
           Component
 Cycle         1.830
 Seasonal     51.089
 Irregular     0.927
 TD & Hol.     2.179
 Others       44.916
 Total       100.941

Combined test in the entire series
 Non parametric tests for stable seasonality
                                                          P.value
   Kruskall-Wallis test                                      0.000
   Test for the presence of seasonality assuming stability   0.000
   Evolutive seasonality test                                0.014
 
 Identifiable seasonality present

Residual seasonality tests
                                      P.value
 qs test on sa                          0.924
 qs test on i                           0.643
 f-test on sa (seasonal dummies)        0.671
 f-test on i (seasonal dummies)         0.453
 Residual seasonality (entire series)   0.415
 Residual seasonality (last 3 years)    0.954
 f-test on sa (td)                      0.091
 f-test on i (td)                       0.333


Additional output variables
y <- mysa$final$series[,"y"]
# De façon équivalente :
y <- get_ts(mysa)
sa <- mysa$final$series[,"sa"]
plot(y)
lines(sa, col = "red")

# ou on peut directement utiliser les fonctions de RJDemetra :
plot(mysa, first_date = 2000, #Pour n'afficher le graphique qu'à partir de 200
     type_chart = "sa-trend" # Pour faire le graphique avec y, sa et tendance
)

Pour des graphiques ggplot2, on peut également utiliser le package ggdemetra :

library(ggdemetra)
# y <- ggdemetra::raw(mysa)
# sa <- ggdemetra::seasonaladj(mysa)
p_sa <- 
    ggplot(data = ggdemetra::ts2df(y), 
           mapping = aes(x = date, y = y)) +
    geom_line(color = "#F0B400") +
    labs(title = "Y, Sa, Trend",
         x = NULL, y = NULL) +
    geom_sa(component = "y_f", linetype = 2,
            spec = x13_spec(mysa), frequency = 12,
            color = "#F0B400") + 
    geom_sa(component = "sa", color = "#155692") +
    geom_sa(component = "sa_f", color = "#155692", linetype = 2)+ 
    geom_sa(component = "t", color = "#1E6C0B") +
    geom_sa(component = "t_f", color = "#1E6C0B", linetype = 2) + 
    theme_bw()
Exercice

Modifier le modèle précédent pour enlever l’effet graduel de Pâques.

spec_sans_easter <- x13_spec(mysa,
                             easter.enabled = FALSE)
mysa2 <- x13(ipi_c_eu[, "FR"], spec_sans_easter)
mysa2$regarima
y = regression model + arima (2, 1, 1, 0, 1, 1)
Log-transformation: yes
Coefficients:
          Estimate Std. Error
Phi(1)     0.05032      0.118
Phi(2)     0.09575      0.082
Theta(1)  -0.55559      0.110
BTheta(1) -0.73033      0.039

              Estimate Std. Error
Week days     0.007179      0.000
AO (5-2011)   0.124297      0.018
LS (11-2008) -0.086926      0.017
LS (1-2009)  -0.071005      0.017


Residual standard error: 0.02133 on 338 degrees of freedom
Log likelihood =   838, aic =  1548 aicc =  1548, bic(corrected for length) = -7.561
Exercice

Calculer les p-valeurs associées au modèle Reg-ARIMA de la précédente spécification.

Récupérer le modèle Reg-ARIMA et utiliser la fonction summary().

reg_sum <- summary(mysa2$regarima)
reg_sum
y = regression model + arima (2, 1, 1, 0, 1, 1)

Model: RegARIMA - X13
Estimation span: from 1-1990 to 12-2019
Log-transformation: yes
Regression model: no mean, trading days effect(1), no leap year effect, no Easter effect, outliers(3)

Coefficients:
ARIMA: 
          Estimate Std. Error  T-stat Pr(>|t|)    
Phi(1)     0.05032    0.11801   0.426    0.670    
Phi(2)     0.09575    0.08224   1.164    0.245    
Theta(1)  -0.55559    0.10956  -5.071 6.45e-07 ***
BTheta(1) -0.73033    0.03938 -18.545  < 2e-16 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Regression model: 
               Estimate Std. Error T-stat Pr(>|t|)    
Week days     0.0071788  0.0003171 22.637  < 2e-16 ***
AO (5-2011)   0.1242970  0.0180390  6.890 2.63e-11 ***
LS (11-2008) -0.0869263  0.0171661 -5.064 6.69e-07 ***
LS (1-2009)  -0.0710049  0.0171449 -4.141 4.34e-05 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1


Residual standard error: 0.02133 on 338 degrees of freedom
Log likelihood =   838, aic =  1548, aicc =  1548, bic(corrected for length) = -7.561

On peut récupérer ces valeurs en exploitant l’objet reg_sum :

reg_sum$coefficients
$arima
             Estimate Std. Error     T-stat     Pr(>|t|)
Phi(1)     0.05031546 0.11801332   0.426354 6.701142e-01
Phi(2)     0.09575287 0.08224054   1.164303 2.451013e-01
Theta(1)  -0.55559277 0.10955864  -5.071191 6.449100e-07
BTheta(1) -0.73033107 0.03938071 -18.545401 0.000000e+00

$regression
                 Estimate   Std. Error    T-stat     Pr(>|t|)
Week days     0.007178836 0.0003171301 22.636876 0.000000e+00
AO (5-2011)   0.124296961 0.0180390210  6.890449 2.628564e-11
LS (11-2008) -0.086926347 0.0171661398 -5.063826 6.685537e-07
LS (1-2009)  -0.071004892 0.0171448840 -4.141462 4.338403e-05

$fixed_out
NULL

$fixed_var
NULL

1.2 Créer un workspace

Dans cette partie nous allons créer un workspace depuis R. Pour cela les fonctions qui peuvent être utilisées sont new_workspace(), load_workspace(), new_multiprocessing(), add_sa_item(), save_workspace(), compute(), get_object(), get_name(), get_ts() ou count().

Exercice

Créer un workspace qui va contenir une série désaisonnalisée selon 3 spécifications différentes.

jws <- new_workspace()
new_multiprocessing(jws, "MP-1")
add_sa_item(jws, "MP-1", mysa, "X13 avec Pâques")
add_sa_item(jws, "MP-1", mysa2, "X13 sans Pâques")
add_sa_item(jws, "MP-1", tramoseats(ipi_c_eu[, "FR"]), "TRAMO-SEATS")
save_workspace(jws, "mon_premier_workspace.xml")
Exercice

Importer le workspace précédent et récupérer :

  • Le nom du premier multi-processing
  • Le nombre de modèles dans ce premier multi-processing
  • L’ensemble des séries brutes
  • Le 2ème modèle
jws <- load_workspace("mon_premier_workspace.xml")
compute(jws)
count(jws) # Nombre de multiprocessing
[1] 1
jmp1 <- get_object(jws, 1) # Le premier multiprocessing
get_name(jmp1)
[1] "MP-1"
count(jmp1)
[1] 3
all_y <- get_ts(jmp1) # toutes les séries brutes
model2 <- get_object(jmp1, 2) # On récupère l'objet associé au 2ème modèle
get_model(model2, jws)
RegARIMA
y = regression model + arima (2, 1, 1, 0, 1, 1)
Log-transformation: yes
Coefficients:
          Estimate Std. Error
Phi(1)     0.05032      0.118
Phi(2)     0.09575      0.082
Theta(1)  -0.55559      0.110
BTheta(1) -0.73033      0.039

              Estimate Std. Error
Week days     0.007179      0.000
AO (5-2011)   0.124297      0.018
LS (11-2008) -0.086926      0.017
LS (1-2009)  -0.071005      0.017


Residual standard error: 0.02133 on 338 degrees of freedom
Log likelihood =   838, aic =  1548 aicc =  1548, bic(corrected for length) = -7.561



Decomposition
Monitoring and Quality Assessment Statistics:
      M stats
M(1)    0.080
M(2)    0.044
M(3)    0.925
M(4)    0.324
M(5)    1.018
M(6)    0.172
M(7)    0.075
M(8)    0.207
M(9)    0.069
M(10)   0.229
M(11)   0.207
Q       0.311
Q-M2    0.344

Final filters: 
Seasonal filter:  3x5
Trend filter:  13 terms Henderson moving average


Final
Last observed values
             y       sa        t         s         i
Jan 2019 103.9 104.9497 104.8872 0.9899977 1.0005966
Feb 2019 101.9 106.2130 105.1277 0.9593928 1.0103234
Mar 2019 111.0 104.7070 105.2207 1.0601007 0.9951182
Apr 2019 107.4 105.0688 105.1200 1.0221871 0.9995130
May 2019 105.5 108.7078 104.9337 0.9704918 1.0359658
Jun 2019 105.8 101.5038 104.6881 1.0423256 0.9695826
Jul 2019 110.1 105.4918 104.3766 1.0436828 1.0106850
Aug 2019  78.7 102.6361 104.0097 0.7667871 0.9867928
Sep 2019 108.5 104.0439 103.5870 1.0428287 1.0044106
Oct 2019 116.8 104.5857 103.1442 1.1167879 1.0139754
Nov 2019 103.8 101.7786 102.7722 1.0198610 0.9903319
Dec 2019  97.7 101.5883 102.5735 0.9617252 0.9903948

Forecasts:
               y_f     sa_f      t_f       s_f       i_f
Jan 2020 101.86190 102.9804 102.5981 0.9891385 1.0037265
Feb 2020 100.79119 103.1440 102.8002 0.9771887 1.0033443
Mar 2020 111.43956 102.5781 103.0224 1.0863878 0.9956869
Apr 2020 105.54428 103.3790 103.2028 1.0209447 1.0017079
May 2020  96.34392 104.3346 103.2155 0.9234131 1.0108422
Jun 2020 112.45079 102.5347 103.0496 1.0967099 0.9950029
Jul 2020 106.92345 102.3231 102.8558 1.0449593 0.9948203
Aug 2020  76.61765 102.5744 102.7208 0.7469472 0.9985747
Sep 2020 110.47115 103.3476 102.6644 1.0689277 1.0066554
Oct 2020 111.53872 102.2010 102.6987 1.0913668 0.9951530
Nov 2020 104.52886 102.5479 102.7719 1.0193176 0.9978206
Dec 2020 101.73775 103.4319 102.9459 0.9836210 1.0047204


Diagnostics
Relative contribution of the components to the stationary
portion of the variance in the original series,
after the removal of the long term trend
 Trend computed by Hodrick-Prescott filter (cycle length = 8.0 years)
           Component
 Cycle         1.984
 Seasonal     62.978
 Irregular     0.998
 TD & Hol.     2.273
 Others       33.676
 Total       101.909

Combined test in the entire series
 Non parametric tests for stable seasonality
                                                          P.value
   Kruskall-Wallis test                                      0.000
   Test for the presence of seasonality assuming stability   0.000
   Evolutive seasonality test                                0.429
 
 Identifiable seasonality present

Residual seasonality tests
                                      P.value
 qs test on sa                          1.000
 qs test on i                           0.965
 f-test on sa (seasonal dummies)        0.625
 f-test on i (seasonal dummies)         0.474
 Residual seasonality (entire series)   0.779
 Residual seasonality (last 3 years)    0.878
 f-test on sa (td)                      0.076
 f-test on i (td)                       0.385


Additional output variables

1.3 Manipuler les objets Java

L’objectif de cette partie est de manipuler la fonction jx13() pour gagner en temps de calcul.

Exercice

Créer un modèle à partir de la fonction jx13() et la spécification sans effet graduel de pâques crée dans la section 1.1.

myjsa <- jx13(ipi_c_eu[, "FR"], spec_sans_easter)
get_indicators(myjsa, "sa")
$sa
           Jan       Feb       Mar       Apr       May       Jun       Jul
1990  93.27297  96.29529  94.47992  93.73334  93.94507  92.83948  94.28898
1991  93.61809  93.12145  92.56800  92.23966  88.02740  94.55282  92.48219
1992  91.62891  91.43164  92.38484  92.19526  91.08704  90.59578  89.51814
1993  87.30634  87.99736  86.85921  86.97894  86.99296  87.29234  86.76534
1994  88.69622  87.77278  88.37550  89.30617  91.27332  91.48108  90.65425
1995  95.09916  94.52731  93.92731  93.90941  92.37904  92.04264  93.77960
1996  93.06670  91.54362  94.26169  93.30918  91.32878  94.96354  94.26156
1997  93.06742  95.67044  95.13059 101.85852  96.43204  98.44794  98.23010
1998 102.13804 103.25275 102.01932 102.89690 103.62209 101.81934 103.60992
1999 103.71964 102.64444 103.76759 104.04665 106.13300 106.82584 105.69365
2000 110.38789 107.38428 109.60527 110.17270 114.61835 106.35539 110.67086
2001 112.45025 113.34404 113.88119 110.73914 111.61677 111.49774 112.13527
2002 110.26696 110.99285 111.13280 111.58735 107.54579 111.87216 110.92070
2003 108.72641 109.77205 109.75157 109.85823 105.81090 105.36978 108.56247
2004 109.12622 109.43552 109.52322 110.42603 111.47708 112.70525 111.38248
2005 114.72877 111.07384 106.23729 114.40231 113.83685 111.14636 109.33782
2006 112.84431 110.57123 111.88826 111.79247 114.34733 112.62975 111.48879
2007 112.53191 114.29580 114.23700 113.06787 112.93424 115.43386 116.32302
2008 114.06045 113.94480 111.07963 118.83142 111.92185 110.96502 111.54674
2009  93.60983  93.32624  92.17978  92.82197  94.61843  94.29186  94.42552
2010  95.51379  96.12075  98.46494  99.01596 102.81820 100.22951  99.30088
2011 104.88512 104.72507 103.63569 102.25128 116.10802  97.55663 101.60353
2012 102.83064  98.48657 101.60323  99.95552  99.30289 100.25687 101.48753
2013  97.45647  98.45690  97.93752 100.35136  99.79029 100.14504 100.08302
2014  97.89835 100.00657  98.78845  99.56271  98.29814  97.26484  98.38400
2015  98.27255  99.53811  99.50288  99.07983  99.28072 101.49169  96.66780
2016 101.32397  99.22467  96.95194 102.97678 105.03035 100.37169  96.68839
2017 102.72859 101.41668 102.63069 100.60801 106.58076 101.54655 100.67684
2018 103.19713 103.14126 103.70513 103.35321 102.04279 105.47625 105.16178
2019 104.94974 106.21301 104.70704 105.06883 108.70777 101.50380 105.49182
           Aug       Sep       Oct       Nov       Dec
1990  92.81532  93.42471  93.21965  92.25071  90.06120
1991  91.72254  92.21808  91.96428  92.02285  90.46195
1992  91.95223  89.69252  89.81277  90.02832  88.33305
1993  87.30196  87.03081  87.21300  84.45288  87.46656
1994  91.37247  91.19741  91.98386  92.78677  95.17153
1995  92.38821  93.99473  92.85706  93.36671  94.38969
1996  93.75684  94.14529  93.72433  93.79062  93.43034
1997 101.81145  99.64928 101.86478 100.90774 101.70138
1998 103.62658 103.60456 103.09096 104.46065 102.99964
1999 106.07083 107.12621 108.29389 107.17738 110.10121
2000 111.01860 110.92744 111.42763 113.10692 114.51591
2001 115.50617 111.31601 110.65583 110.20538 108.13496
2002 112.75124 109.55689 109.18036 109.90888 106.38982
2003 107.51695 108.10694 110.01758 109.50133 108.16481
2004 110.31329 111.75188 112.44658 108.46909 111.73847
2005 110.00423 112.59573 109.23754 111.65979 113.81029
2006 112.62602 113.48750 112.76155 112.55067 113.59525
2007 115.82841 112.91803 115.49734 113.69439 112.28630
2008 109.36385 109.09852 108.03726 100.63716  99.86454
2009  96.91866  97.94748  97.85917  97.60414  96.07651
2010  99.54513 100.43464 100.75477  98.83302 102.97020
2011 101.60617 101.45456 102.21445 102.59847 103.85017
2012 101.83135  99.96547  98.23490  98.64730  97.23489
2013  96.84385  97.95614  99.86479  98.42704  97.79728
2014  96.12996  99.28707  98.34888  97.27581  99.01780
2015 102.05354 101.35873 100.66252 100.76102 100.45864
2016 101.62022 100.64268  98.76241 100.83623 102.82647
2017 103.73341 103.82246 105.29816 106.59767 105.53516
2018 105.03058 103.00706 104.49092 104.89205 103.66170
2019 102.63605 104.04393 104.58566 101.77858 101.58827
Exercice

Toujours avec la même spécification, extraire les révisions de la séries désaisonnalisée du point de janvier 2005 (i.e. : série désaisonnalisée lorsqu’on a les données jusqu’en janvier 2005, puis jusqu’en février 2005, etc.).

Récupérer l’ensemble des dates de fin d’estimation avec la fonction window(time(ipi_c_eu[, "FR"]), start = 2005).

dates <- as.numeric(window(time(ipi_c_eu[, "FR"]), start = 2005))
estimations <- sapply(dates, function(last_date_estimation){
    myjsa <- jx13(window(ipi_c_eu[, "FR"], end = last_date_estimation), spec_sans_easter)
    sa <- get_indicators(myjsa, "sa")$sa
    window(sa, start = 2005, end = 2005) # Pour ne récupérer que la valeur en 2005
})
estimations <- ts(estimations, start = 2005, frequency = 12)
plot(estimations)

2 RJDemetra v3

2.1 Créer une specification

Les modèles autour de la méthode X-13ARIMA peuvent être estimés avec rjd3x13::x13(), rjd3x13::regarima()) et rjd3x13::x11().

Les spécifications peuvent être crées avec les fonctions rjd3x13::regarima_spec, rjd3x13::spec_x11() ou rjd3x13::x13_spec() et peuvent être modifiées par :

  • Pour le préajustement : rjd3toolkit::set_arima(), rjd3toolkit::set_automodel(), rjd3toolkit::set_basic(), rjd3toolkit::set_easter(), rjd3toolkit::set_estimate(), rjd3toolkit::set_outlier(), rjd3toolkit::set_tradingdays(), rjd3toolkit::set_transform(), rjd3toolkit::add_outlier(), rjd3toolkit::remove_outlier(), rjd3toolkit::add_ramp(), rjd3toolkit::remove_ramp(), rjd3toolkit::add_usrdefvar() ;

  • Pour la décomposition : rjd3x13::set_x11() ;

  • Pour le benchmarking : rjd3toolkit::set_benchmarking().

Exercice

Faire la désaisonnalisation d’une série avec X-13 avec la spécification suivante :

  • détection automatique du schéma de décomposition, des outliers et du modèle ARIMA ;
  • une correction des jours ouvrables “working days” et un effet graduel de Pâques.

Faire ensuite un graphique avec la série brute et la série désaisonnalisée.

Utiliser la spécification RSA4c pour la désaisonnalisation.

Si le modèle créé s’appelle sa_jd3, regarder les valeurs de sa_jd3$result$final et rjd3toolkit::sa_decomposition(sa_jd3).

sa_jd3 <- rjd3x13::x13(ipi_c_eu[, "FR"], "rsa4")
sa_jd3
RegARIMA
Log-transformation: yes 
SARIMA model:  (2,1,1) (0,1,1)

Coefficients
          Estimate Std. Error  T-stat
phi(1)     0.02269    0.10769   0.211
phi(2)     0.15463    0.07367   2.099
theta(1)  -0.52470    0.09974  -5.261
btheta(1) -0.70977    0.04324 -16.415

Regression model:
                  Estimate Std. Error T-stat
td               0.0070856  0.0002973 23.830
easter          -0.0218293  0.0041947 -5.204
LS (2008-11-01) -0.0851379  0.0169175 -5.033
LS (2009-01-01) -0.0720198  0.0168966 -4.262
AO (2011-05-01)  0.1267563  0.0170943  7.415
Number of observations:  360 
Number of effective observations:  347 
Number of parameters:  10 

Loglikelihood:  850.2857 
Adjusted loglikelihood:  -752.4816 

Standard error of the regression (ML estimate):  0.02060681 
AIC:  1524.963 
AICC:  1525.618 
BIC:  1563.456 


Decomposition
Monitoring and Quality Assessment Statistics: 
    M stats
m1    0.087
m2    0.049
m3    0.934
m4    0.713
m5    0.965
m6    0.212
m7    0.075
m8    0.198
m9    0.070
m10   0.209
m11   0.195
q     0.340
qm2   0.376

Final filters: 
Seasonal filter:  
Trend filter: 13 terms Henderson moving average

Diagnostics
Relative contribution of the components to the stationary
portion of the variance in the original series,
after the removal of the long term trend (in %)

           Component
 cycle         8.148
 seasonal     53.817
 irregular     0.338
 calendar      2.074
 others       30.402
 total        94.779

Residual seasonality tests
                P.value
 seas.ftest.i     0.122
 seas.ftest.sa    0.486
 seas.qstest.i    0.600
 seas.qstest.sa   0.788
 td.ftest.i       0.383
 td.ftest.sa      0.420


Final
Last values
         series       sa    trend      seas       irr
Jan 2019  103.9 104.9459 104.8892 0.9900340 1.0005406
Feb 2019  101.9 106.1674 105.2359 0.9598051 1.0088514
Mar 2019  111.0 104.5521 105.4496 1.0616713 0.9914890
Apr 2019  107.4 105.9774 105.4357 1.0134233 1.0051380
May 2019  105.5 108.7623 105.2678 0.9700057 1.0331958
Jun 2019  105.8 101.1875 104.9615 1.0455836 0.9640439
Jul 2019  110.1 105.3177 104.5343 1.0454088 1.0074942
Aug 2019   78.7 102.5250 104.0408 0.7676178 0.9854307
Sep 2019  108.5 104.0024 103.5185 1.0432448 1.0046754
Oct 2019  116.8 104.5360 103.0474 1.1173179 1.0144465
Nov 2019  103.8 101.7273 102.7183 1.0203752 0.9903519
Dec 2019   97.7 101.6177 102.5972 0.9614465 0.9904527
y <- sa_jd3$result$preadjust$a1
sa <- sa_jd3$result$final$d11final
# ou :
decomp <- rjd3toolkit::sa_decomposition(sa_jd3)
y <- decomp$series
sa <- decomp$sa
# ou on peut directement utiliser les fonctions de rjd3x13 :
plot(sa_jd3, first_date = 2000 #Pour n'afficher le graphique qu'à partir de 200
)

Pour des graphiques ggplot2, on peut également utiliser le package ggdemetra3.

Exercice

Modifier le modèle précédent pour enlever l’effet graduel de Pâques.

spec_sans_easter_v3 <- 
    sa_jd3$estimation_spec |> 
    rjd3toolkit::set_easter(enabled = FALSE)
sa2_jd3 <- rjd3x13::x13(ipi_c_eu[, "FR"], spec_sans_easter_v3)
sa2_jd3$result$preprocessing
Log-transformation: yes 
SARIMA model:  (2,1,1) (0,1,1)

Coefficients
          Estimate Std. Error  T-stat
phi(1)     0.05029    0.10689   0.470
phi(2)     0.09574    0.07647   1.252
theta(1)  -0.55562    0.09576  -5.802
btheta(1) -0.73033    0.03972 -18.387

Regression model:
                  Estimate Std. Error T-stat
td               0.0071788  0.0003171 22.637
LS (2008-11-01) -0.0869273  0.0171661 -5.064
LS (2009-01-01) -0.0710048  0.0171448 -4.141
AO (2011-05-01)  0.1242974  0.0180391  6.890
Number of observations:  360 
Number of effective observations:  347 
Number of parameters:  9 

Loglikelihood:  838.005 
Adjusted loglikelihood:  -764.7623 

Standard error of the regression (ML estimate):  0.02132569 
AIC:  1547.525 
AICC:  1548.059 
BIC:  1582.168
Exercice

Calculer les p-valeurs associées au modèle Reg-ARIMA de la précédente spécification.

Récupérer le modèle Reg-ARIMA et utiliser la fonction summary().

summary(sa2_jd3$result$preprocessing)
Log-transformation: yes 
SARIMA model:  (2,1,1) (0,1,1)

Coefficients
          Estimate Std. Error  T-stat Pr(>|t|)    
phi(1)     0.05029    0.10689   0.470    0.638    
phi(2)     0.09574    0.07647   1.252    0.211    
theta(1)  -0.55562    0.09576  -5.802  1.5e-08 ***
btheta(1) -0.73033    0.03972 -18.387  < 2e-16 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Regression model:
                  Estimate Std. Error T-stat Pr(>|t|)    
td               0.0071788  0.0003171 22.637  < 2e-16 ***
LS (2008-11-01) -0.0869273  0.0171661 -5.064 6.76e-07 ***
LS (2009-01-01) -0.0710048  0.0171448 -4.141 4.36e-05 ***
AO (2011-05-01)  0.1242974  0.0180391  6.890 2.72e-11 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Number of observations:  360 , Number of effective observations:  347 , Number of parameters:  9 
Loglikelihood:  838.005, Adjusted loglikelihood:  -764.7623
Standard error of the regression (ML estimate):  0.02132569 
AIC:  1547.525 , AICc:  1548.059 , BIC:  1582.168

2.2 Créer un workspace

Dans cette partie nous allons créer un workspace depuis R. Pour cela les fonctions qui peuvent être utilisées sont rjdemetra3::.jws_new(), rjdemetra3::read_workspace(), rjdemetra3::.jws_sap_new(), rjdemetra3::add_sa_item(), rjdemetra3::save_workspace().

Exercice

Créer un workspace qui va contenir une série désaisonnalisée selon 3 spécifications différentes.

jws <- rjdemetra3::.jws_new()
jsap <- rjdemetra3::.jws_sap_new(jws, "MP-1")
rjdemetra3::add_sa_item(jsap, "X13 avec Pâques", sa_jd3)
rjdemetra3::add_sa_item(jsap, "X13 sans Pâques", sa2_jd3)
rjdemetra3::add_sa_item(jsap, "TRAMO-SEATS", y, rjd3tramoseats::spec_tramoseats())
rjdemetra3::save_workspace(jws, "ws_v3.xml")
Exercice

Importer le workspace précédent et récupérer :

  • Le nom du premier multi-processing
  • Le nombre de modèles dans ce premier multi-processing
  • L’ensemble des séries brutes
  • Le 2ème modèle
# charge tous les modèles du workspace :
jws <- rjdemetra3::.jws_load("ws_v3.xml")
ws <- rjdemetra3::read_workspace(jws)
names(ws$processing)[1]
[1] "MP-1"
length(ws$processing[[1]]) # Nombre de multiprocessing
[1] 3

Autre façon de faire similaire à RJDemetra :

# charge tous les modèles du workspace :
jws <- rjdemetra3::.jws_load("ws_v3.xml")
rjdemetra3::.jws_compute(jws)
rjdemetra3::.jws_sap_count(jws) # Nombre de multiprocessing
[1] 1
jsap1 <- rjdemetra3::.jws_sap(jws, 1) # Le premier multiprocessing
rjdemetra3::.jsap_name(jsap1)
[1] "MP-1"
rjdemetra3::.jsap_sa_count(jsap1)
[1] 3
model2 <- rjdemetra3::.jsap_sa(jsap1, 2) # On récupère l'objet associé au 2ème modèle
rjdemetra3::.jsa_read(model2)
$ts
$name
[1] "X13 sans Pâques"

$moniker
$source
[1] ""

$id
[1] "f15592ce-2d48-4450-89ca-5a2f3a8f4644"

attr(,"class")
[1] "JD3_TSMONIKER"

$metadata
NULL

$data
       Jan   Feb   Mar   Apr   May   Jun   Jul   Aug   Sep   Oct   Nov   Dec
1990  92.1  92.3 102.1  93.0  93.3 100.8  92.9  66.7  95.8 105.0  96.7  89.2
1991  92.5  89.2  97.4  93.8  87.5 100.3  93.4  64.3  96.9 103.5  94.0  92.1
1992  90.7  89.0  99.4  93.7  86.1 101.3  90.4  62.9  96.6  98.4  91.9  92.6
1993  82.3  84.0  95.6  88.3  82.2  97.9  85.5  61.3  93.7  93.0  88.3  92.1
1994  83.6  83.7  97.0  88.3  88.3 102.9  87.3  65.9  98.2  98.0  96.8  98.0
1995  91.8  90.1 102.9  90.4  91.6 103.7  90.6  66.8  98.7 101.4  97.2  94.8
1996  92.0  91.1  98.1  94.3  90.5 101.8  96.1  66.3  98.9 105.0  95.0  96.0
1997  91.9  91.3  99.1 102.8  93.2 108.2 100.4  70.5 107.3 114.1  99.6 106.7
1998  98.2  98.7 109.3 103.7  97.6 114.7 106.1  72.1 111.5 112.6 105.6 107.4
1999  97.2  98.3 114.5 104.8  99.9 120.2 105.7  76.1 115.2 115.1 111.1 114.0
2000 103.4 107.5 121.7 105.7 113.1 119.4 108.1  82.0 116.4 121.3 117.2 111.9
2001 110.7 108.9 124.0 109.3 109.8 121.9 112.4  85.5 114.1 123.4 114.2 104.9
2002 108.4 106.7 118.5 113.4 105.6 119.2 113.9  81.4 115.6 121.7 111.0 105.2
2003 106.9 105.4 117.1 112.0 101.5 115.2 111.2  75.7 117.5 122.4 107.8 109.3
2004 104.7 106.7 122.8 112.7 104.5 126.5 111.1  79.7 121.9 118.8 112.2 112.6
2005 107.6 106.3 118.8 113.7 109.7 125.0 106.4  81.7 123.0 115.1 115.5 111.6
2006 108.8 105.9 124.8 108.0 113.1 126.7 108.7  84.1 121.0 121.5 116.6 108.2
2007 111.5 109.6 124.0 111.7 111.7 126.6 116.6  87.1 117.3 127.2 118.0 106.5
2008 113.2 114.4 117.5 120.2 107.6 121.4 115.1  78.8 118.9 118.8  99.5  99.2
2009  90.7  90.1 100.0  93.9  88.3 105.5  97.9  70.3 106.5 104.7  99.0  95.2
2010  90.3  93.1 109.5 100.4  95.5 111.8 100.8  74.5 109.0 105.0 102.7 101.9
2011  99.0 101.6 115.3 101.6 110.1 108.5 101.0  78.3 110.0 106.4 106.3 100.2
2012  99.3  99.9 110.3  99.8  96.1 108.5 103.8  78.8 102.9 107.6 101.9  91.5
2013  96.3  95.7 103.8 103.3  96.2 105.4 105.2  73.4 103.3 109.3  99.0  94.5
2014  96.7  97.0 104.8 102.8  92.3 104.7 103.2  71.4 107.2 107.8  95.6  98.3
2015  94.6  96.1 108.4 102.4  90.9 111.8 101.1  76.1 109.2 107.8 101.9  99.9
2016  95.2  99.6 108.4 103.5  98.7 110.4  95.9  79.9 108.1 103.6 105.0  99.7
2017  99.1  97.3 114.7  98.3 102.9 111.5  99.8  81.7 108.5 113.7 111.3  99.6
2018 102.2  98.9 112.9 103.2  98.8 112.8 106.9  82.7 104.8 116.3 109.7  97.5
2019 103.9 101.9 111.0 107.4 105.5 105.8 110.1  78.7 108.5 116.8 103.8  97.7

attr(,"class")
[1] "JD3_TS"

$domainSpec
Specification

Series
Serie span: All 
Preliminary Check: Yes

Estimate
Model span: All 

Tolerance: 1e-07

Transformation
Function: AUTO
AIC difference: -2
Adjust: NONE

Regression
Calendar regressor: WorkingDays
with Leap Year: Yes
AutoAdjust: TRUE
Test: REMOVE

Easter: No

Pre-specified outliers: 0
Ramps: No

Outliers
Detection span: All 
Outliers type: 
    - AO, critical value : 0 (Auto)
    - LS, critical value : 0 (Auto)
    - TC, critical value : 0 (Auto)
TC rate: 0.7 (Auto)
Method: ADDONE (Auto)

ARIMA
SARIMA model:  (0,1,1) (0,1,1)

Coefficients
          Estimate      Type
theta(1)         0 UNDEFINED
btheta(1)        0 UNDEFINED

Specification X11
Seasonal component: Yes
Length of the Henderson filter: 0
Seasonal filter: FILTER_MSR
Boundaries used for extreme values correction : 
     lower_sigma:  1.5 
     upper_sigma:  2.5
Nb of forecasts: -1
Nb of backcasts: 0
Calendar sigma: NONE

Benchmarking
Is enabled: No

$estimationSpec
Specification

Series
Serie span: All 
Preliminary Check: Yes

Estimate
Model span: All 

Tolerance: 1e-07

Transformation
Function: AUTO
AIC difference: -2
Adjust: NONE

Regression
Calendar regressor: WorkingDays
with Leap Year: Yes
AutoAdjust: TRUE
Test: REMOVE

Easter: No

Pre-specified outliers: 0
Ramps: No

Outliers
Detection span: All 
Outliers type: 
    - AO, critical value : 0 (Auto)
    - LS, critical value : 0 (Auto)
    - TC, critical value : 0 (Auto)
TC rate: 0.7 (Auto)
Method: ADDONE (Auto)

ARIMA
SARIMA model:  (0,1,1) (0,1,1)

Coefficients
          Estimate      Type
theta(1)         0 UNDEFINED
btheta(1)        0 UNDEFINED

Specification X11
Seasonal component: Yes
Length of the Henderson filter: 0
Seasonal filter: FILTER_MSR
Boundaries used for extreme values correction : 
     lower_sigma:  1.5 
     upper_sigma:  2.5
Nb of forecasts: -1
Nb of backcasts: 0
Calendar sigma: NONE

Benchmarking
Is enabled: No

$pointSpec
Specification

Series
Serie span: All 
Preliminary Check: Yes

Estimate
Model span: All 

Tolerance: 1e-07

Transformation
Function: LOG
AIC difference: -2
Adjust: LEAPYEAR

Regression
Calendar regressor: WorkingDays
with Leap Year: No
AutoAdjust: FALSE
Test: NO

Easter: No

Pre-specified outliers: 3
    - LS (2008-11-01), coefficient: -0.0869273387886007 (ESTIMATED)
    - LS (2009-01-01), coefficient: -0.0710047923357392 (ESTIMATED)
    - AO (2011-05-01), coefficient: 0.124297351091159 (ESTIMATED)
Ramps: No

Outliers
Is enabled: No

ARIMA
SARIMA model:  (2,1,1) (0,1,1)

Coefficients
             Estimate      Type
phi(1)     0.05029197 ESTIMATED
phi(2)     0.09573828 ESTIMATED
theta(1)  -0.55562180 ESTIMATED
btheta(1) -0.73032859 ESTIMATED

Specification X11
Seasonal component: Yes
Length of the Henderson filter: 0
Seasonal filter: FILTER_MSR
Boundaries used for extreme values correction : 
     lower_sigma:  1.5 
     upper_sigma:  2.5
Nb of forecasts: -1
Nb of backcasts: 0
Calendar sigma: NONE

Benchmarking
Is enabled: No

$results
RegARIMA
Log-transformation: yes 
SARIMA model:  (2,1,1) (0,1,1)

Coefficients
          Estimate Std. Error  T-stat
phi(1)     0.05029    0.10689   0.470
phi(2)     0.09574    0.07647   1.252
theta(1)  -0.55562    0.09576  -5.802
btheta(1) -0.73033    0.03972 -18.387

Regression model:
                  Estimate Std. Error T-stat
td               0.0071788  0.0003171 22.637
LS (2008-11-01) -0.0869273  0.0171661 -5.064
LS (2009-01-01) -0.0710048  0.0171448 -4.141
AO (2011-05-01)  0.1242974  0.0180391  6.890
Number of observations:  360 
Number of effective observations:  347 
Number of parameters:  9 

Loglikelihood:  838.005 
Adjusted loglikelihood:  -764.7623 

Standard error of the regression (ML estimate):  0.02132569 
AIC:  1547.525 
AICC:  1548.059 
BIC:  1582.168 


Decomposition
Monitoring and Quality Assessment Statistics: 
    M stats
m1    0.080
m2    0.044
m3    0.925
m4    0.324
m5    1.018
m6    0.172
m7    0.075
m8    0.207
m9    0.069
m10   0.229
m11   0.207
q     0.311
qm2   0.344

Final filters: 
Seasonal filter:  
Trend filter: 13 terms Henderson moving average

Diagnostics
Relative contribution of the components to the stationary
portion of the variance in the original series,
after the removal of the long term trend (in %)

           Component
 cycle         8.120
 seasonal     54.041
 irregular     0.305
 calendar      1.946
 others       30.441
 total        94.853

Residual seasonality tests
                P.value
 seas.ftest.i     0.175
 seas.ftest.sa    0.566
 seas.qstest.i    0.956
 seas.qstest.sa   1.000
 td.ftest.i       0.281
 td.ftest.sa      0.279


Final
Last values
         series       sa    trend      seas       irr
Jan 2019  103.9 104.9497 104.8872 0.9899977 1.0005966
Feb 2019  101.9 106.2130 105.1277 0.9593928 1.0103234
Mar 2019  111.0 104.7070 105.2207 1.0601008 0.9951181
Apr 2019  107.4 105.0688 105.1200 1.0221872 0.9995130
May 2019  105.5 108.7078 104.9337 0.9704918 1.0359659
Jun 2019  105.8 101.5038 104.6881 1.0423255 0.9695828
Jul 2019  110.1 105.4918 104.3766 1.0436829 1.0106850
Aug 2019   78.7 102.6360 104.0097 0.7667871 0.9867928
Sep 2019  108.5 104.0439 103.5870 1.0428287 1.0044107
Oct 2019  116.8 104.5857 103.1442 1.1167879 1.0139754
Nov 2019  103.8 101.7786 102.7722 1.0198610 0.9903319
Dec 2019   97.7 101.5883 102.5735 0.9617251 0.9903948
# Pour extraire toutes les séries brutes il faut faire une boucle :
all_y <- lapply(
    lapply(
        seq_len(rjdemetra3::.jsap_sa_count(jsap1)),
        rjdemetra3::.jsap_sa, jsap = jsap1
    ),
    rjdemetra3::get_raw_data
)

2.3 Manipuler les objets Java

La manipulation des objets Java se fait de la même façon qu’avec RJDemetra : jx13() pour estimer les modèles, rjd3toolkit::dictionary() pour connaitre l’ensemble des objets exportables et rjd3toolkit::result() et rjd3toolkit::user_defined() pour exporter des objets.

Exercice

Créer un modèle à partir de la fonction jx13() et la spécification sans effet graduel de pâques calculée dans la section 2.1.

myjsa <- rjd3x13::jx13(ipi_c_eu[, "FR"], spec_sans_easter_v3)
rjd3toolkit::result(myjsa, "sa")
           Jan       Feb       Mar       Apr       May       Jun       Jul
1990  93.27297  96.29529  94.47992  93.73334  93.94507  92.83948  94.28898
1991  93.61809  93.12145  92.56800  92.23966  88.02740  94.55282  92.48219
1992  91.62891  91.43164  92.38484  92.19526  91.08704  90.59578  89.51814
1993  87.30634  87.99736  86.85921  86.97894  86.99296  87.29234  86.76534
1994  88.69622  87.77278  88.37550  89.30617  91.27332  91.48108  90.65425
1995  95.09916  94.52731  93.92731  93.90941  92.37904  92.04264  93.77960
1996  93.06670  91.54362  94.26169  93.30918  91.32878  94.96354  94.26156
1997  93.06742  95.67044  95.13059 101.85852  96.43204  98.44794  98.23010
1998 102.13804 103.25275 102.01932 102.89690 103.62209 101.81934 103.60992
1999 103.71964 102.64444 103.76759 104.04665 106.13300 106.82584 105.69365
2000 110.38789 107.38428 109.60527 110.17270 114.61835 106.35539 110.67086
2001 112.45025 113.34404 113.88119 110.73914 111.61677 111.49774 112.13527
2002 110.26696 110.99285 111.13280 111.58735 107.54579 111.87216 110.92070
2003 108.72641 109.77205 109.75157 109.85823 105.81090 105.36978 108.56247
2004 109.12622 109.43552 109.52322 110.42603 111.47708 112.70525 111.38248
2005 114.72876 111.07384 106.23729 114.40231 113.83685 111.14636 109.33782
2006 112.84430 110.57123 111.88827 111.79247 114.34733 112.62975 111.48879
2007 112.53191 114.29580 114.23701 113.06787 112.93424 115.43386 116.32303
2008 114.06045 113.94479 111.07964 118.83142 111.92185 110.96502 111.54674
2009  93.60983  93.32624  92.17979  92.82197  94.61843  94.29186  94.42552
2010  95.51378  96.12075  98.46495  99.01596 102.81821 100.22951  99.30088
2011 104.88511 104.72507 103.63569 102.25128 116.10803  97.55663 101.60353
2012 102.83064  98.48657 101.60323  99.95552  99.30289 100.25687 101.48753
2013  97.45647  98.45690  97.93752 100.35136  99.79029 100.14504 100.08302
2014  97.89835 100.00657  98.78845  99.56271  98.29814  97.26484  98.38400
2015  98.27255  99.53811  99.50288  99.07983  99.28072 101.49169  96.66780
2016 101.32397  99.22467  96.95194 102.97678 105.03035 100.37169  96.68839
2017 102.72859 101.41668 102.63068 100.60800 106.58076 101.54656 100.67684
2018 103.19713 103.14126 103.70512 103.35321 102.04279 105.47626 105.16178
2019 104.94974 106.21301 104.70703 105.06882 108.70777 101.50380 105.49182
           Aug       Sep       Oct       Nov       Dec
1990  92.81532  93.42471  93.21965  92.25071  90.06120
1991  91.72254  92.21808  91.96428  92.02285  90.46195
1992  91.95223  89.69252  89.81277  90.02832  88.33305
1993  87.30196  87.03081  87.21300  84.45288  87.46656
1994  91.37247  91.19741  91.98386  92.78677  95.17153
1995  92.38821  93.99473  92.85706  93.36671  94.38969
1996  93.75684  94.14529  93.72433  93.79062  93.43034
1997 101.81145  99.64928 101.86478 100.90774 101.70138
1998 103.62658 103.60456 103.09096 104.46065 102.99964
1999 106.07083 107.12621 108.29389 107.17738 110.10121
2000 111.01860 110.92744 111.42763 113.10692 114.51591
2001 115.50617 111.31601 110.65583 110.20538 108.13496
2002 112.75124 109.55689 109.18036 109.90888 106.38982
2003 107.51695 108.10694 110.01758 109.50133 108.16481
2004 110.31329 111.75189 112.44658 108.46909 111.73847
2005 110.00423 112.59573 109.23755 111.65978 113.81028
2006 112.62602 113.48751 112.76156 112.55066 113.59524
2007 115.82842 112.91804 115.49735 113.69438 112.28629
2008 109.36385 109.09853 108.03727 100.63715  99.86453
2009  96.91866  97.94749  97.85918  97.60413  96.07650
2010  99.54513 100.43465 100.75478  98.83301 102.97020
2011 101.60617 101.45457 102.21445 102.59847 103.85016
2012 101.83135  99.96547  98.23490  98.64730  97.23489
2013  96.84385  97.95614  99.86479  98.42704  97.79728
2014  96.12996  99.28707  98.34888  97.27581  99.01780
2015 102.05354 101.35873 100.66252 100.76102 100.45864
2016 101.62022 100.64268  98.76241 100.83623 102.82647
2017 103.73341 103.82246 105.29816 106.59767 105.53517
2018 105.03057 103.00707 104.49092 104.89205 103.66171
2019 102.63605 104.04393 104.58566 101.77858 101.58828

Pour les révisions, la fonction rjd3x13::x13_revisions() facilite grandement la tâche. L’historique des révisions peut s’exporter à partir de trois paramètres :

  • data_ids qui permet d’exporter des statistiques ;

  • ts_ids qui permet d’exporter des estimations d’une composante à une certaine date ;

  • cmp_ids qui permet l’ensemble des estimations d’une composante à une ensemble de dates.

data_ids <- list(
    # Export du coefficient du premier régresseur jours ouvrables
    list(start = "2005-01-01", id = "regression.td(1)"))
ts_ids <- list(
    # Export de l'historique des estimations de la séries SA de janv 2010 à partir de janvier 2010
    list(period = "2010-01-01", start = "2010-01-01", id = "sa"))
cmp_ids <- list(
    # Export de l'ensemble des tendances estimées entre janv 2010 et dec 2014
    list(start = "2010-01-01", end = "2014-12-01", id = "t"))
rh <- rjd3x13::x13_revisions(
    sa2_jd3$result$preadjust$a1,
    spec = sa2_jd3$result_spec, 
    data_ids, ts_ids, cmp_ids)
plot(rh$data$`regression.td(1)`,
     ylab = "Coef", main = "Coef td")

plot(rh$series$sa,
     ylab = NULL, main = "SA de janv 2010")

ncol(rh$components$t) # nombre de séries exportées
[1] 60