Mixed Models and Spline-Based Smoothing

Martin Boer, Wageningen University and Research

Abstract: The LMMsolver R package provides an efficient approach for fitting (Generalized) Linear Mixed Models using sparse linear algebra. A key advantage is its seamless integration of B-splines within the mixed model framework, enabling computationally efficient smoothing for spatial and spatio-temporal data. Compared to traditional methods, LMMsolver significantly reduces computational time while maintaining scalability. We illustrate its performance by applying two-dimensional smoothing to Sea Surface Temperature data, demonstrating the benefits of sparse mixed model formulations for large datasets.