Multiple time series • pickmdl3

The pickmdl3 package includes functionality to support an orderly production process when using the seasonal adjustment methods of rjd3x13. The function x13_text_frame is a wrapper function that seasonally adjusts a multivariate time series object with the x13_pickmdlfunction, based on a data frame with model specification settings.

The object rti_mts contains three time series:


rti_mts <- do.call(cbind,pickmdl3::pickmdl_data("norwegian_rti"))

head(rti_mts)
#>          rti_472 rti_476 rti_4751
#> Jan 2014    61.1    71.9     64.5
#> Feb 2014    62.8    62.9     60.8
#> Mar 2014    68.6    69.8     68.9
#> Apr 2014    76.5    72.2     65.2
#> May 2014    78.4    71.9     69.1
#> Jun 2014    79.3    79.9     71.5

We can use the function make_param_fileto define a data frame with specifications. The first input to this function is the multivariate time series object that is to be adjusted. Further input are the specification settings that are used to define the specification with which to adjust the series in question. With this we mean all specification settings that are available in rjd3toolkit, rjd3x13 or the options of x13_pickmdl. With the exception of the initial specification setting spec, the reference to settings in rjd3tookitand rjd3x13 need to be given on the format function__setting as illustrated below.

spec_file <- make_paramfile(rti_mts, spec = "rsa3",
              set_outlier__outliers.type = c("LS","AO"),
              set_transform__fun = "Log",
              automdl.enabled = TRUE)

spec_file 
#>       name   spec set_outlier__outliers.type set_transform__fun automdl.enabled
#> 1  rti_472 "rsa3"              c("LS", "AO")              "Log"            TRUE
#> 2  rti_476 "rsa3"              c("LS", "AO")              "Log"            TRUE
#> 3 rti_4751 "rsa3"              c("LS", "AO")              "Log"            TRUE
#>                                                                                         userdefined
#> 1 c("diagnostics.seas-si-combined", "diagnostics.seas-sa-friedman", "residuals.independence.value")
#> 2 c("diagnostics.seas-si-combined", "diagnostics.seas-sa-friedman", "residuals.independence.value")
#> 3 c("diagnostics.seas-si-combined", "diagnostics.seas-sa-friedman", "residuals.independence.value")

The make_paramfile function constructs a data frame with the same specifications for all series. To adjust specifications settings for individual series, use the function edit_constraints and do the necessary adjustments in the shiny application that pops up.


spec_file <- edit_constraints(spec_file)


spec_file$set_outlier__outliers.type[2] <- NA
spec_file$spec[3] <- "\"rsa5c\""
spec_file$automdl.enabled[1] <- "FALSE"

Resulting in a frame with individual settings for each series. For example:

spec_file
#>       name    spec set_outlier__outliers.type set_transform__fun
#> 1  rti_472  "rsa3"              c("LS", "AO")              "Log"
#> 2  rti_476  "rsa3"                       <NA>              "Log"
#> 3 rti_4751 "rsa5c"              c("LS", "AO")              "Log"
#>   automdl.enabled
#> 1           FALSE
#> 2            TRUE
#> 3            TRUE
#>                                                                                         userdefined
#> 1 c("diagnostics.seas-si-combined", "diagnostics.seas-sa-friedman", "residuals.independence.value")
#> 2 c("diagnostics.seas-si-combined", "diagnostics.seas-sa-friedman", "residuals.independence.value")
#> 3 c("diagnostics.seas-si-combined", "diagnostics.seas-sa-friedman", "residuals.independence.value")

Note that a blank field means that the setting is not used for the series in question.

Now, to do the seasonal adjustment of the multivariate time series object based on this data frame, we use the function x13_text_frame Output of the function is a list of sa output objects for each time series in the mts.


sa_mult <- pickmdl3::x13_text_frame(text_frame= spec_file,ts = "rti_mts")
#> Warning in crit_selection(crit_tab_i, star = 0, when_star = when_star_here): No
#> model is ok according to criteria

To evaluate indivudual series:


sa_mult$rti_472
#> Serie span: All 
#> 
#> Model: X-13
#> Log-transformation: yes 
#> SARIMA model: (0,1,1) (0,1,1)
#> 
#> SARIMA coefficients:
#>  theta(1) btheta(1) 
#>   -0.7788   -0.9998 
#> 
#> Regression model:
#> AO (2021-03-01) 
#>          0.2224 
#> 
#>  Seasonal filter: FILTER_S3X5;  Trend filter: H-23 terms
#>  M-Statistics: q Good (0.590); q-m2 Good (0.644)
#>  QS test on SA: Good (1.000);  F-test on SA: Good (1.000)
#> 
#> For a more detailed output, use the 'summary()' function.
sa_mult$rti_476
#> Serie span: All 
#> 
#> Model: X-13
#> Log-transformation: yes 
#> SARIMA model: (0,1,1) (0,1,1)
#> 
#> SARIMA coefficients:
#>  theta(1) btheta(1) 
#>   -0.8032   -0.3312 
#> 
#> Regression model:
#> LS (2020-02-01) AO (2020-03-01) 
#>          0.1854         -0.3411 
#> 
#>  Seasonal filter: FILTER_S3X5;  Trend filter: H-23 terms
#>  M-Statistics: q Good (0.671); q-m2 Good (0.724)
#>  QS test on SA: Good (0.534);  F-test on SA: Good (0.760)
#> 
#> For a more detailed output, use the 'summary()' function.
sa_mult$rti_4751
#> Serie span: All 
#> 
#> Model: X-13
#> Log-transformation: yes 
#> SARIMA model: (1,0,2) (0,1,1)
#> 
#> SARIMA coefficients:
#>    phi(1)  theta(1)  theta(2) btheta(1) 
#>   -0.8132   -0.3264   -0.1841   -0.4981 
#> 
#> Regression model:
#>     const       mon       tue       wed       thu       fri       sat    easter 
#>  0.036961 -0.012811  0.006242  0.009563 -0.008628  0.017098  0.010154 -0.070936 
#> 
#>  Seasonal filter: FILTER_S3X5;  Trend filter: H-23 terms
#>  M-Statistics: q Good (0.558); q-m2 Good (0.597)
#>  QS test on SA: Good (0.482);  F-test on SA: Good (0.993)
#> 
#> For a more detailed output, use the 'summary()' function.