|
Abstract:
|
Subject-level heterogeneity may result with biased estimators and wrong statistical conclusion in behavioral and psychiatric studies if not addressed correctly. The commonly applied methods in both observational studies and clinical trials introduce artificial equilibriums among subjects, whereas, they are heterogeneous. Such study results also prevents making suitable inference over the validity of the problem-solving approaches. In this presentation, a two-level regression model will be applied in a design that incorporates subject-level repeated measures in behavioral and psychiatric studies. Studies with subject-level repeated measures tend to be correlated within subject. Even though the correlation seems to imply loss of statistical power, it can be compensated by increasing the within-subject sample size without increasing subjects and, accounting for the intra-subject correlation in the model. This approach aims to map within-subject variation and address the two sources of variation, within and between subjects. Using Monte Carlo simulation, statistical inferences and power analyses under the proposed method will be carried out and compared with other methods.
|
