All Sparse PCA Models Are Wrong, But Some Are Useful. Part I: Computation of Scores, Residuals and Explained Variance
This work addresses a critical computational issue for practitioners using sPCA to improve data interpretation, highlighting limitations in current methods.
The paper identifies that sparse PCA (sPCA) models lack orthogonality in scores and loadings, preventing direct application of traditional PCA methods for computing scores, residuals, and explained variance, and demonstrates through simulations that sPCA algorithms show disparate and limited performance on noise-free, sparse data.
Sparse Principal Component Analysis (sPCA) is a popular matrix factorization approach based on Principal Component Analysis (PCA) that combines variance maximization and sparsity with the ultimate goal of improving data interpretation. When moving from PCA to sPCA, there are a number of implications that the practitioner needs to be aware of. A relevant one is that scores and loadings in sPCA may not be orthogonal. For this reason, the traditional way of computing scores, residuals and variance explained that is used in the classical PCA cannot directly be applied to sPCA models. This also affects how sPCA components should be visualized. In this paper we illustrate this problem both theoretically and numerically using simulations for several state-of-the-art sPCA algorithms, and provide proper computation of the different elements mentioned. We show that sPCA approaches present disparate and limited performance when modeling noise-free, sparse data. In a follow-up paper, we discuss the theoretical properties that lead to this problem.