Modelling of covariance structure in constrained marginal models for longitudinal data

A data-driven method for modelling intra-subject covariance matrix is introduced to constrained marginal models with longitudinal data. A constrained iteratively re-weighted least squares algorithm is presented consequently. Asymptotic properties of the constrained ML estimates, including strong consistency, approximate representation and asymptotic distribution, are given. Real data analysis and simulations are conducted to compare our new approach with classical menu-selection-based modelling technique.