An omnibus noise filter (Q964661)

A noise filtering algorithm is proposed based on flexible least squares (FLS) estimation (i.e., minimization of penalized least squares). A sequential updating procedure is presented for the solution of the minimization problem. The FLS is compared to Kalman filtering and Wiener-Kolmogorov filtering via simulations for stationary and nonstationary short- and long-memory processes with and without structural breaks. An application to some exchange rates data is presented.