PC-Programme zur Zeitreihenanalyse: Datenmanagement, Grafik und univariate Analyseverfahren (SPSS, SYSTAT, STATISTICA, MicroTSP, MESOSAUR)

Abstract

Der vorliegende Artikel vergleicht fünf Software-Programme für Zeitreihenanalysen: drei allgemeine statistische Programme (SPSS, SYSTAT, STATISTICA), eine ökonometrische Anwendung (MicroTSP) und ein speziell für Zeitreihenanalysen entwickeltes Programm (MESOSAUR). Zusätzlich zu den relevanten statistischen Verfahren werden Möglichkeiten der Datentransformation und des Datenmanagements sowie Formen graphischer Aufbereitung dargestellt. Für SPSS und SYSTAT werden unterschiedliche DOS- und Windows-Versionen berücksichtigt. Die EDV-technische Umsetzung (univariater) zeitreihenanalytischer Verfahren wird nach technischen Anforderungen und Kosten beschrieben und bewertet. (pmb)

'This article compares five software-packages for time series analysis: three general statistical packages (SPSS, SYSTAT, STATISTICA), one econometric application (MicroTSP) and one program specifically designed for time series analysis (MESOAUR). In addition to relevant statistical procedures possibilities for data transformation, datahandling and graphical facilities will also be dealt with. For SPSS and SYSTAT differences between DOS- and Windows-versions are noted. The statistical procedures for time series analysis of all three general packages are fairly similar: techniques for smoothing by moving averages, for fitting trend functions, exponential smoothing, spectral analysis and ARIMA-model building. The procedures related to the latter will be discussed in detail. MicroTSP, which is clearly an economic program, lacks some 'classical' methods like spectral analysis and symmetric moving averages. On the other hand some newer (econometric) developments like unit root tests are implemented. Finally the program MESOSAUR will be described, which has unique data-management facilities. The various algorithms of this program even contain frequency filtering. On the other hand one can not recommend ARIMA-procedures as well as the built-in expert system. As a result some specific requirements concerning historical research will be formulated. Finally some further references to other time series analysis software and statistical programming languages will be given.' (author's abstract)