SA Workshop Luxembourg, 6/3/2012
What is JDemetra+ Objectives of JDemetra+ Main features ◦ New core engines ◦ Additional statistical tools Current state Planning ◦ Next steps Short Demo
New implementation of Demetra+, completely in Java, available in open source ◦ Links with Demetra+ Same logic Similar presentation Compatible I/O ◦ Pure Java application: no Fortran code ◦ Open source software EUPL license Code available on JoinUp (
Why JDemetra+? ◦ Maintenance in the long term (leave Fortran) ◦ Extensions Peripheral: quality reports... Methods: New models: structural models... Modifications of existing method: new filters for X11... Extensions of some features: new calendar variables...
Why Java? ◦ Portable, fast, reliable, widely diffused ◦ Standard environment for many institutions Eurostat, ECB... ◦ Many free libraries, tools JFreeChart, NetBeans, Eclipse... Why an open software? ◦ Easier integration in in-house developments ◦ Sharing of knowledge, developments... ◦ Common framework for statistical modules
New core engines ◦ Comparability with the original core engines: > 90% (differences mainly due to other optimization procedures) ◦ Common framework for the pre-processing parts ◦ Easily reusable building blocks Outliers detection... Additional statistical tools ◦ Modeling (SA): Structural models, generalized airline models... ◦ Temporal disaggregation: Chow-Lin, Fernandez, Litterman... ◦ Benchmarking: Denton, Cholette... ◦ State space framework
Current state (March 2012) ◦ Core engines (statistical routines) Tramo-Seats: last improvements are missing X13: quasi complete ◦ Graphical interfaces for the main SA algorithms: quasi complete ◦ TS providers: finished ◦ Output, compatibility with workspaces of Demetra+: in progress (50%) ◦ Additional modules (direct/indirect, quality report, diagnostics): in progress (25%)
Final objective ◦ New modules available for production at the end of Next steps ◦ Core engines (for tests): April 2012 ◦ NetBeans framework: May 2012 ◦ First prototype of JDemetra+: June 2012