Statistical Methods in Music Corpus Studies: Application, Use Cases, and Best Practice Examples

In this chapter, the authors explain that there are two common goals in musical corpus analysis. The first is the description and comparison of musical corpora, the second is to establish relationships between musical structures and extra-musical data, which can refer to metadata of a particular musical piece (genre, style, and period labels, composer and performer attributions, etc.) or to listeners’ perceptions and evaluations. The authors give a brief overview of basic and advanced statistical methods that have been employed in music corpus studies. The chapter covers descriptive statistics and visualizations, feature selection and aggregation using principal component analysis. In addition, random forests and linear regression methods for use in the context of corpus studies are briefly explained, as well as supervised and unsupervised classification techniques. Each topic and method is introduced with a conceptual explanation, suggestions for its application, and usage scenarios from the research literature.