Sparse component analysis for linear mixed models

When seeking for a sparse solution of a linear model, a common technique is the search of a solution with minimum ℓ₁ norm. In this paper, we present a new approach for the case of sparse linear mixed models. We combine the EM algorithm for solving the inverse problem with a decision test that guarantees sparseness by eliminating the statistically null components of the solution. We address its performance by means of simulations and illustrate its use with real radar data demonstrating its potential applications.