Developing academic software for teaching time series analysis : a case study

Abstract

ABSTRACT: The academic training on time series analysis requires not only a sound theoretical background on the methods but also the use of specific academic software to appreciate the methods’ capabilities, limitations and proper applicability. It is desirable for students to program the routines and algorithms by themselves but this is not always feasible, particularly during short courses and workshops in which the interest is to understand the information supplied by several analytical methods. Considering the time constraints and the need to stress the interpretative rather than the computational skills, the authors have developed the software package UATSA (University of Arizona Time Series Analysis) that incorporates many analytical tools commonly used in time series analysis in an organized and sequential manner: exploratory statistics, markovian processes, univariate and multivariate analyses (ARMA models), frequency decomposition algorithms, principal components, canonical correlations and cluster analyses are included within the current version of the package. UATSA is a stand-alone executable file compiled in MATLAB® that has been used in courses of time series analysis in hydrology at the University of Arizona and in several workshops offered by the authors between 2004 and 2007. The software aims to easily illustrate the use of algorithms in the synthesis and decomposition of time series, providing a background to the methods and a visual platform that is user friendly and data extensive. The software has evolved through the years, incorporating suggestions made by students to improve its appearance and widen its scope. The software also has contributed to a shift in teaching dynamics by allowing students and instructors to focus on interpreting and analyzing outcomes rather than just learning the set of mathematical tools.