Abstract:
|
The main advantage of longitudinal studies is that they can distinguish changes over time within individuals (longitudinal effects) from differences among subjects at the start of the study (cross-sectional effects). However, in most studies, longitudinal changes need to be studied after correction for potential important cross-sectional differences between subjects. In this presentation, the effect will be studied of cross-sectional model misspecifications, on the inference for the longitudinal effects, in the context of linear mixed models. For balanced data, it will be shown that the longitudinal component of a well-formulated linear mixed model is orthogonal to the cross-sectional component, implying that inference for longitudinal effects is independent, is not affected by possible misspecifications of cross-sectional effects. For unbalanced data however, both components are no longer completely orthogonal, implying that misspecified cross-sectional models may introduce bias in the statistical inference for the longitudinal effects, which are the parameters of interest. However, simulations and theoretical arguments will be used to show that this bias is small in practice.
|