In this article we consider the problem of sample size determination for two-level mixed-effects logistic regression models for the analysis of longitudinal data. In this case, level 1 represents measurement occasion and level 2 represents subject. We provide an algorithmfor sample size requirements (i.e., power characteristics) for a test of the treatment-by-time interaction and overall time trend that strictly maintains the type I error rate close to the nominal level. The methodology is illustrated both via simulation and an example from a now classic psychiatric randomized clinical trial.
|Number of pages||11|
|Journal||Journal of Applied Statistical Science|
|Publication status||Published - 1 Dec 2011|
- Hierarchical models
- Marginal maximum likelihood
- Power analysis
- Type I error rate