Conceptual

Two-Stage Estimator for High-Dimensional Markov-Switching ODE Processes

A statistical framework and two-stage algorithm for recovering the parameters of high-dimensional Markov-switching ordinary-differential processes with nonlinear additive dynamics from discrete observations: the first stage reconstructs the continuous sample path from samples and the second estimates the switching and drift parameters. The analysis truncates the latent posterior processes and proves statistical-error bounds and a linear-convergence guarantee under beta-mixing conditions, and the method is used to contrast resting-state brain-network transition-rate matrices between ADHD and control groups.