A1018
Title: The mean group estimators for multi-level autoregressive models with intensive longitudinal data
Authors: Kazuhiko Hayakawa - Hiroshima University (Japan)
Boyan Yin - Hiroshima University (Japan) [presenting]
Abstract: The mean group (MG) estimators are proposed to estimate multilevel (vector) autoregressive models with intensive longitudinal data. The MG estimator is originally proposed in econometrics, but is new to the behavioural science. Since the naive MG estimator suffers from the small sample bias problem, jackknife and analytical bias corrections are proposed. It is argued that the MG estimator has several advantages over existing methods, such as restricted maximum likelihood or Bayesian methods in terms of model specification and implementation. Monte Carlo simulation is performed to investigate the performance of the MG estimators and compare them with the existing methods. The simulation results indicate that the bias-corrected MG estimators have superior or comparable performance compared to the existing methods.
