Masters of PCA: The Evolution of Metallica Through the Years with Exploratory Data Analysis. An Invitation to Teaching PCA in a Different Way

Metallica is one of the cornerstones of metal rock music, and the go-to band for starting to explore this genre [1]. It is a band central to any discourse around the multi-faced topic of what is real rock and metal music, but also a very successful powerhouse of musical production and live performances. Principal Component Analysis (PCA, [2]) is, arguably, the cornerstone of chemometrics, and the go-to method for exploratory data analysis. It is a technique central to any course in multivariate data analysis, but also a very efficient tool for compressing data, to the point that many people, especially those from the many branches of informatics, only use it as a compression tool, thus not fully exploiting its potential. In our study we want to investigate the meeting point between Metallica and PCA, to answer a very common research question about the former: “Has Metallica evolved through the years in their studio albums attending to strict music indicators?”. This debate has been always a central point discussion between supporters and critics of the band. The approach considers a careful identification and selection of the data that can be used to describe the musical production (not the recording quality) of the band throughout their studio albums, coupled with the power of PCA, which is simply used to model and extract information from such dataset. The parallel aim of the study, and the crucial message behind this research, is to convey the idea behind PCA in a clear and fresh way, making its use, potentiality and functioning as clear as possible to both the novice (who might be struggling with grasping the full power of this technique) and the expert user (who might be sick and tired of always seeing the usual chemical dataset used for explaining PCA). With the idea of eventually shaping it into a tutorial, the whole data analysis pipeline will be described together with tricks and suggestions on how to structure and conduct it in potentially any other real-world scenario, showing the strengths and weaknesses of the data collection procedure, the database construction, the application of PCA and the interpretation of the results. Disclaimer: Why Metallica? That’s an easy one. Because we love it!