Large Scale Applied Time Series Analysis with Program TSW (TRAMO-SEATS for Windows)

The demostration will center on the application of program TSW to a large set of monthly time series. TSW is a Windows interface of updated versions of programs TRAMO (Time series Regression with Arima noise, Missing values, and Outliers) and SEATS (Signal Extraction in ARIMA Time Series). The program estimates a general regression-ARIMA model, and computes forecasts and interpolators for possibly nonstationary series, with any sequence of missing observations, and in the presence of outliers. The program contains an option for automatic model identification, automatic detection and correction of several types of outliers, and for pretesting and estimation of Calendar-based effects. Several types of intervention or regression variables can also be included. Next, the program estimates and forecasts the trend, seasonal, calendar, transitory, and noise components in the series, using signal extraction techniques applied to ARIMA models. The program contains a part on diagnosis and on inference, and an analysis of the properties of the estimators and of the estimation and forecasting errors. The last part of the output is oriented towards its use in short term economic policy and monitoring. TRAMO contains an extension (TERROR, or Tramo for ERRORs) to the problem of quality control of data in large data bases of time series; SEATS can be applied for estimation of long-term trends and (business) cycles. The programs can efficiently and reliably handle, in an entirely automatic manner, applications to sets of many thousand series. They are already being used intensively in research, data producing agencies, policy making institutions, and business. (Perhaps the most widely used application is Seasonal Adjustment.) They are freely available, together with documentation, at the Bank of Spain web site (www.bde.es).