Time Varying Parameter Estimation Scheme for a Linear Stochastic Differential Equation

Abstract

In this work, an attempt is made to estimate time varying parameters in a linear stochastic differential equation. By defining $m_{k}$ as the local admissible sample/data observation size at time $t_{k}$, parameters and state at time $t_{k}$ are estimated using past data on interval $[t_{k-m_{k}+1}, t_{k}]$. We show that the parameter estimates at each time $t_{k}$ converge in probability to the true value of the parameters being estimated. A numerical simulation is presented by applying the local lagged adapted generalized method of moments (LLGMM) method to the stochastic differential models governing prices of energy commodities and stock price processes.