Estimating Linear Mixed-effects State Space Model Based on Disturbance Smoothing

TL;DR

This paper extends linear mixed-effects state space models to correlated individuals, proposing new algorithms for parameter and state estimation, validated through numerical studies demonstrating their effectiveness.

Contribution

It introduces explicit recursive formulas for EM and score algorithms and proposes the MKF-KS algorithm for state estimation without prior knowledge of random effects.

Findings

Explicit recursive formulas for EM maximizer

Score vector formulas show ML equals moment estimation

Numerical validation confirms algorithm efficacy

Abstract

We extend the linear mixed-effects state model to accommodate the correlated individuals and investigate its parameter and state estimation based on disturbance smoothing in this paper. For parameter estimation, EM and score based algorithms are considered. Intermediate quantity of EM algorithm is investigated firstly from which the explicit recursive formulas for the maximizer of the intermediate quantity are derived out for two given models. As for score based algorithms, explicit formulas for the score vector are achieved from which it is shown that the maximum likelihood estimation is equivalent to moment estimation. For state estimation we advocate it should be carried out without assuming the random effects being known in advance especially when the longitudinal observations are sparse. To this end an algorithm named mixture Kalman filter with kernel smoothing (MKF-KS) is proposed. Numerical studies are carried out to investigate the proposed algorithms which validate the efficacy of the proposed inference approaches.