--- title: "Plug-in JVS - EN" output: rmarkdown::html_vignette vignette: > %\VignetteIndexEntry{Plug_in_JVS_EN} %\VignetteEngine{knitr::rmarkdown} %\VignetteEncoding{UTF-8} --- ```{r, include = FALSE} knitr::opts_chunk$set( collapse = TRUE, comment = "#>" ) ``` ```{r setup} library("JDCruncheR") ``` ## Introduction This vignette explains how to create a JVS Quality Report with the {JDCruncheR} package. This only works for v3. ## Definition of the JVS QR The JVS (Job Vacancy Survey) plug-in is an add-on for the JDemetra+ software. ### Version 2 Version 2 is available on GitHub [Sace2015](https://github.com/Sace2015): https://github.com/Sace2015/EurostatQR. You will find the [plug-in documentation](https://github.com/Sace2015/EurostatQR/raw/refs/heads/master/2018-10-26%20A%20PlugIn%20For%20Eurostat%20Job%20Vacancy%20Survey.docx) on the same repository. ### Version 3 In version 3, the plug-in is hosted on [jdemetra](https://github.com/jdemetra/)’s GitHub, in the [jdplus-incubator](https://github.com/jdemetra/jdplus-incubator/) repository. To download the latest version, go to the [Releases](https://github.com/jdemetra/jdplus-incubator/releases) section, then download the plug-in (named jdplus-eurostat-desktop-plugin). ### Description The plug-in contains 30 columns: - **Series**: Name of the series - **Method**: Seasonal adjustment method used (Tramo-Seats or X13) and the specification - **Period**: Periodicity of the series (12 for monthly and 4 for quarterly) - **Nobs**: Number of observations - **Start**: Starting date - **End**: Ending date - **Adjustment**: Kind of adjustment performed: NSA(non seasonally and calendar adjusted), CA (calendar adjusted) or SCA (seasonally and calendar adjusted) - **Presence of Seasonality in the Raw Series**: Result (Yes, No, Uncertain) of a a combined test based on 3 tests: the Fisher test, the Kruskal-Wallis test and the Ljung-Box test. - **Presence of TD effects**: Result (Yes or No) of the JD+ Fisher joint test. - **Log-Transformation**: Decomposition model. If "Yes": A logarithmic transformation is done by Tramo-Seats for the complete adjustment; A logarithmic transformation is done by X13 for the Reg-Arima part and a multiplicative model is used for the decomposition per se. - **ARIMA Model**: Orders (p,d,q)(P,D,Q) of the ARIMA model - **LeapYear**: Presence of a Leap Year effect (Yes or No) - **MovingHoliday**: Presence of an Easter effect (Yes or No) - **NbTD**: Number of trading-day regressors (including the Leap Year regressor) - **Noutliers**: Total number of outliers - **Outlier1**, **Outlier2**, **Outlier3**: The 3 main outliers. The importance is measured by the absolute value of the t-statistic associated with the outlier. - **Residual Seasonality in SA Series (F-test)**: Presence of residual seasonality in the SCA series. The test used here is a Fisher test. The test is done at the 5% level. - **Residual TD Effect**: Presence of residual trading-day effect in the SCA series. The test used here is a Fisher test based on the default Gregorian trading-day regressors. The test is done at the 5% level. - **Q-Stat (for X13)**: Final Q statistics (available for X13 only) - **Final Henderson Filter**: Length of the Henderson filter used in the final estimation of the trend-cycle in table D12 (available for X13 only). - **Stage 2 Henderson Filter**: Length of the Henderson filter used in the provisional estimation of the trend-cycle in table D6 (available for X13 only). - **Seasonal Filter**: Seasonal filter used in the final estimation of the seasonal component in table D10 (available for X13-adjusted series). - **Irregular Standard-Deviation**: Standard deviation of the irregular defined as (SA - Trend) - **Quality (for TS)**: JDemetra+ global quality assessment - **Max-Adj**: Indicator of the "size" of the seasonal and calendar adjustment. $$ 100 \times max(\left | \frac{Y_t - SA_t}{Y_t}\right |) $$ - **Autocorrelation of order 1 of the SA series**: Reports the autocorrelation of order 1 of the series $SA_t - SA_{t-1}$ - **Normality test (P-value)**: P-value of the Normal test to check if the autocorrelation is significant. - **Negative and significant autocorrelation**: A warning is issued when the autocorrelation is negative and significant. ## Creation of the output files To create a JVS Quality Report, you need to use the JDemetra+ cruncher in version 3. First, if you do not have downloaded the JDemetra+ cruncher, you can download it from GitHub : https://github.com/jdemetra/jdplus-main/releases. The file you should download starts with "jwsacruncher-standalone". To call the cruncher, you can use the {rjwsacruncher} package with specific options. The goal of this step is to generate a demetra_m file containing a lot of informations on series to study, and also several files containing the components of the series. To generate those files, you need to use the cruncher with the following options: ```{r, echo = TRUE, eval = FALSE} library("rjwsacruncher") options( is_cruncher_v3 = TRUE, default_matrix_item = c( "span.start", "span.end", "log", "span.n", "regression.nout", "regression.ntd", "m-statistics.m7", "decomposition.seasonal-filters", "decomposition.trend-filter", "decomposition.d7-trend-filter", "quality.summary", "regression.lp", "regression.leaster", "diagnostics.seas-sa-ac1:3", "diagnostics.seas-sa-ac1", "diagnostics.seas-lin-combined", "regression.out(*)", "regression.td-ftest:3", "residuals.lb:3", "diagnostics.td-sa-last:2", "diagnostics.seas-sa-f:2", "arima.p", "arima.d", "arima.q", "arima.bp", "arima.bd", "arima.bq", "m-statistics.q", "m-statistics.q-m2" ), default_tsmatrix_series = c( "decomposition.y_cmp", "decomposition.s_cmp", "decomposition.sa_cmp", "decomposition.t_cmp" ) ) cruncher_and_param( workspace = "path/to/the/workspace/file.xml", cruncher_bin_directory = "path/to/the/cruncher/v3/bin/directory", csv_layout = "vtable", short_column_headers = FALSE, refreshall = FALSE, v3 = TRUE, delete_existing_file = TRUE, policy = "None" ) ``` At this step, you should have generated 5 files. ## Creation of report JVS Once the cruncher has generated the correct outputs, the JVS Quality Report can be generated. ```{r, echo = TRUE, eval = TRUE} # Path leading to the directory containing the needed files dir_path <- system.file( "extdata", "WS/WS_world/Output/SAProcessing-1", package = "JDCruncheR" ) JVS <- extract_JVS(dir = dir_path) ``` ```{r, echo = FALSE, eval = TRUE} rownames(JVS) <- NULL knitr::kable(JVS[, 1L:7L]) ``` The JVS object is a dataframe containing the quality report. To save it, you need to export it. ## Export To export the JVS Quality Report, you need to use the writing function. The report can be exported as a .csv file: ```{r, echo = TRUE, eval = FALSE} write(JVS, format = "csv", export_dir = "path/to/the/export/directory", overwrite = TRUE) ``` The report can be exported as a .xlsx file: ```{r, echo = TRUE, eval = FALSE} write(JVS, format = "xlsx", export_dir = "path/to/the/export/directory", overwrite = TRUE) ``` ## Comparison: v3 / R ### X The column is called "Series" in R. ### Method The names of the specifications are simply "X13" or "Tramo-Seats" in R, as there is no detail regarding the type of specification. ### Period, nobs, Start No differences. ### End In v3, the dates correspond to the last day of the period, whereas in R they correspond to the first day. ### Adjustment No differences. ### Tests For the tests: - Presence of Seasonality in the RAW Series - Presence of TD effects - Residual Seasonality in SA Series F-test - Residual TD Effect Very slight differences due to different thresholds between the interface and the cruncher. ### Log transformation, ARIMA Model, Leapyear, MovingHoliday, NbTD, Noutliers, Outlier1, Outlier2, Outlier3, Q-stat for X13, Stage 2 Henderson Filter, Final Henderson Filter, Seasonal.Filter No differences. ### Irregular Standard Deviation There are slight differences between v3 (which uses the correct number of degrees of freedom) and the R code, which uses `df = 1` (with the `sd()` function). In future versions of the JVS plug-in, the results will be identical. ### Quality for TS, Max adj, Autocorrelation of order 1 of the SA series No differences. ### Normal test The test is carried out using base R and may show slight differences (to within 10**4). ### Negative and significant autocorrelation No differences.