BATS

Transcribed from the printed issue; not yet reviewed. Read it in the PDF of the issue, page 36

In the last issue of Vector, the program BATS - Bayesian Analysis of Time Series - was wrongly attributed. BATS is produced by and copyrighted to Andy Pole, Mike West and Jeff Harrison (see the entry in the Product Guide).

For those readers who are unfamiliar with BATS, I thought it might be useful to include a little detail on the software and the authors.

Andy Pole is visiting professor at, and Mike West is Director of, the Institute of Statistics and Decision Sciences at Duke University in North Carolina, USA; Jeff Harrison is professor of statistics at Warwick University.

BATS is an interactive system for modelling, analysing, and forecasting univariate time series using dynamic linear models. A wide range of dynamic models may be selected, including components for polynomial trend; seasonal patterns; exogenous variable regression; and variance laws. The analysis features include reference or user-defined initialisations, forecast monitoring, subjective intervention, step ahead forecasting, and smoothing. Non-stationarity, missing data, and outlying observations are routinely handled by the Bayesian paradigm. BATS is graphically oriented. Tabulated results may be written to text files.

BATS was originally developed in APL*PLUS/PC but has now been redeveloped in C (with the same ‘look and feel’) to improve performance. Both versions are available from the authors but support and future development will be restricted to the C versions.