Emery Schubert, University of New South Wales

Music, musical phenomena and music perception consist of times-series because they are intrinsically temporally dependent.  This intrinsic temporal nature is not acknowledged in much music perception research, and means that the assumptions of the time-series are often violated because traditional statistical modelling is applied.  The lack of acknowledgement also means that the richness of data, particularly when the music signal and response are changing, can be lost. This unit covers some basic time-series techniques for the analysis of data stemming from time dependent music and music responses. Included in the unit are when to use time-series analysis, methods of collecting continuous response data, methods of investigating the nature of a time-series and serial correlation, ways of modelling dependent and independent variables, and methods of diagnosis. Techniques such as cross-correlations and ARIMA modelling will be introduced.  Although some of the mathematics in time series analysis is complex, this unit will focus on conceptual issues and on applying simple solutions and models in practical situations.