Abstract

Analyzing curves is a common task in speech research. For example, pitch contours are analyzed (and also modified) in intonation research. Usually a few quantitative descriptors of such curves are extracted first(duration, peaks, max velocity, etc.) then ordinary statistical analyses are applied to those descriptors. Functional Data Analysis (FDA) is a set of modern techniques that allow you to perform statistical analyses directly on the entire set of curves. It provides both qualitative insight in the form of graphical display of results, as well as quantitative output in the well known form of p-values, percentage of explained variance, etc.

An oral presentation will provide a global view of FDA tools mainly by showing applied examples on intonation. A hands-on session will follow, in which the free R software package for FDA will be illustrated.