A1615
Title: A sparse Kalman filter: A non-recursive approach
Authors: Jan Bruha - CNB (Czech Republic) [presenting]
Abstract: An algorithm is proposed to filter states and shocks in a state-space model under sparsity. Many interesting economic models (such as linearized DSGE models, trend-cyclical VARs, time-varying VARs, dynamic factor models) can be cast into linear state-space models. Under the conventional Kalman filter, which is essentially a recursive OLS algorithm, all shocks are estimated to be non-zero. Sparsity may be beneficial for statistical efficiency and we argue that for some applications, the sparse solution is natural. The sparsity of filtered shocks is achieved by an elastic-net penalty. The algorithm that can be straightforwardly adapted for non-convex penalties or to achieve robustness to outliers