General Mixed Linear Model
Recall from Chapter 6 that the assumed mean for the response of the ith subject at occasion j was
E Y Xij ij( ) = ′β (7.1)
where: ′Xij is a 1 by p vector of covariate values
β is a p by 1 vector of unknown regression parameters
In addition, it was assumed that the responses for subject i is a multivariate normal with covariance matrix Σi for i = 1,2,…,N. Various patterns for the covariance matrix were described, and Bayesian techniques were employed to estimate the unknown parameters. Several patterns for the covariance matrix were included: unstructured, autoregressive, compound symmetry, and Toeplitz.