Prerequisite
The JAVA modules need the Java SDK 6.01 or higher.
The latest version can be downloaded from the WEB site of Sun . The Java run-time 6.02 for Windows can be downloaded from here.
NBB Libraries
The file JNbb_bin.zip contains the statistical libraries developed by the NBB (NbbTools.jar). Beside some auxiliary libraries, it also contains three small applications that can be launched as stand-alone modules.
- JX11.jar is the NBB's implementation of the X11 seasonal adjustment algorithm (see the US Bureau of the Census for the original program).
- JTramoSeats.jar is the NBB's implementation of the TramoSeats seasonal adjustment algorithm (see the Bank of Spain for the original program).
- JGAirline.jar provides a GUI for the Generalized Airline model.
Those applications rely, for some graphical presentations, on the JFreeChart library. That library is distributed under the terms of the GNU Lesser General Public Licence (LGPL), which permits use in proprietary applications.
The source code for those applications is provided in the file JNbbAPI.zip; it should provide some information on the way of using the libraries.
Finally, JNbb_bin.zip also contains some data that can be used by the applications. The time series are stored in a simple xml format; so they can be visualized through standard xml/html browsers and new input files can be easily generated. The java class that implements the xml serialization (be.nbb.timeseries.simplets.xml.Formatter) is also in JNbbAPI.zip.
Some common uses of the libraries are explained in the case studies.
|
The applications should not be considered as production tools, but as small prototypes. Their aim is simply to illustrate the capabilities of the NBB’s algorithmic libraries, not only for the making of dynamic graphical interfaces but also for the automatic processing of a lot of series. Advanced users are encouraged to modify the existing code and to customize the applications for their own needs. |
The three applications are built on the same scheme; the following explanations apply to each of them.
At the starting of the application, a small set of time series is provided (2 series). They are displayed in the tree panel on the left side of the main windows, under the node “Data”. The set of the time series is split in several sub-sets (the initial data set contains only one collection). More interesting sets of time series are provided in the subdirectory “./data/*.xml” provided with the software. They can be loaded by means of the main menu: “File->Open WS…”.
The algorithm is executed by means of a double click on some final node in the time series tree. A window with graphics on the main results is automatically opened for the first processing. Subsequent processing will reuse that window.
Each processing uses the current specifications of the algorithm. Those specifications can be modified through the main menu (“Algorithm->Specifications…”), where Algorithm stands for X11, TramoSeats or Generalized airline. Changes of the specifications are directly applied on any visible results (see below).
Several output windows can be opened for each algorithm (“Algorithm->Results->…”). Beside the main graphics, tables, periodograms, quality tests and detailed text information are available. It should be noted that they are just some (quick and dirty) examples of possible output.
New instances of each algorithm can be generated by the main menu (“File->New analysis”). By default, a new analysis becomes the active one; any operation described above applies on that algorithm. The different algorithms are displayed in the second part of the tree window. Some basic actions can be applied on each of them by means of the local menu (available through the right bottom of the mouse). That feature can be useful to compare results on the same series based on different specifications.
The item “Algorithm->Batch processing” of the main menu starts the processing of all the series of the data set. The same specifications, initialized at the beginning of the process, are applied on all the series. Those specifications only applied on that batch processing. Several useful points should be noted: every batch processing is executed in a separated thread; that means that several processing can run at the same time and/or that the user can continue its interactive analysis; moreover, a batch processing can be stopped at any time; the results are stored in an xml file that can be easily examined; finally, some feedback can be given to the user during the processing (see SGAirline).
The speed of the libraries and their robustness - some series are really ugly - are well illustrated by the different batch processing.