Description
We apply dynamics graphics to time series analysis.
In this field many crude graphical devices are available such
as time-plots and histograms but in general it is difficult to interact
with them. The software we propose focuses on analysis of distributional
characteristics of the series, such as level, variability and stationarity,
existence of components
such as trend component and seasonal component, and finally relation
between these latters. To achieve this result we use dynamically some
graphical tool based on summarised data. These tools are parallel boxplots
(Hoaglin et al., 1982), quantile traces (D'Esposito, 1994), and
scatter-plot matrices (Becker et al., 1988).
The tools we present allow to choose interactively a power
transformation which promotes stationarity and/or additivity of the
component.
The XLisp-Stat (V2.1, R3, A6) used to develop the code is the one
for Ms-Windows v3.1. The preliminary version of the code can be retrieved
via anonymous ftp server ftp.unipg.it (141.250.1.7) on Internet.
The self extracting file dt4ets.exe is located in the directory
/pub/stat/xlispstat/contributions. Click
here to get the software.
The preliminary version of the software have been presented
at the software section of the 11th CompStat held in Vienna (AT) from
22 to 26 aug 1994.
References
Becker R.A., W.S. Cleveland, A.R. Wilks (1987) Dynamic Graphics
for Data Analysis - Stat. Sci., 2, 355-383.
D'Esposito M.R. (1994) Exploratory Graphical Analysis of Time
Series Data - CompStat '94, Short Communication.
Hoaglin D.C., F. Mosteller, J.W. Tukey (1982) Understanding Robust and
Exploratory Data Analysis - Wiley, New York
Project Investigators
Maria Rosaria D'Esposito & Stefano M. Pagnotta
Dipartimento di Matematica e Statistica
Universita' di Napoli Federico II
via Cinthia, I-80126 Naples, Italy.
Contact person
Stefano M. Pagnotta
pagnotta@ds.cised.unina.it