Van Oirbeek, RobinWaseige, JulienJulienWaseige2025-05-142025-05-142025-05-142024https://hdl.handle.net/2078.2/39662In the context of variable selection for regression, many methods exist. One way to handle many variables is to reduce the dimensionality through principal component analysis and then perform the regression onto the components. Sparse versions of principal component analysis have been developed to enhance this approach further. Indeed, principal component analysis fails when the data structure is more complex, or rather, sparse. While sparse principal component methods share some common features and a common history, they also differ in various ways. This thesis reviews these methods, in terms of performance, and type. Additionally, these techniques will be compared to determine which one is best suited for different scenarios. An interesting alternative would be to use sparse principal components analysis as a variable selection method: this idea is explored throughout the thesis on a logistical model, with notable results for the GPower method.RegressionPCASPCAPCRSPCRGLMMLASSOSVDtext::thesis::master thesisthesis:49218