Comparison of Nonlinear Filters: Monte-Carlo Experiments

Abstract

In this chapter, based on the criteria of BIAS (bias) and RMSE (root mean squared error), we examine the nonlinear filters introduced and developed in the previous chapters. Monte-Carlo experiments are performed in Section 5.2. There, each nonlinear filter is compared using various types of nonlinear functions. One set of data y_t and α_t for t=l,...,T is artificially simulated and, given y_t, each filtering estimate of α_t is compared with the artificially simulated α_t. This procedure is performed 1000 times (i.e., 1000 sets of data are generated) and BIAS and RMSE between the estimated α_t and the simulated one are computed for each time t.