A Self-Organizing State Space Model and Simplex Initial Distribution Search

Abstract

This paper proposes a method to seek initial distributions of parameters for a self-organizing state space model proposed by Kitagawa (1998). Our method is based on the simplex Nelder-Mead algorithm for solving nonlinear and discontinuous optimization problems. We show the effectiveness of our method by applying it to a linear Gaussian model, a linear non-Gaussian model, a nonlinear Gaussian model, and a stochastic volatility model.

Keywords: Self-organizing state space model, Monte Carlo particle filter, Parameter estimation, Simplex Nelder-Mead algorithm

JEL Classification: C22