ABSTRACT
This chapter discusses two extensions to the bivariate dual change score (BDCS) model, including: alternative specifications of the initial conditions (ICs) in the BDCS model; and the feasibility of incorporating stochasticity at both the latent change as well as the measurement level. It illustrates the implementation and assesses the performance of these models using simulated data sets. A longitudinal data set with measures of children's reading and arithmetic skills from kindergarten years through middle school was used to demonstrate the use of these BDCS models and corresponding differences in inferential results. The chapter extends the BDCS model to include both process noises and measurement errors, and evaluated the estimation quality of the stochastic BDCS model, especially when coupled with different degrees of misspecification in the IC structure. The original BDCS model assumes that the underlying latent dynamic processes are deterministic or do not show within-person uncertainties at the latent level.
