|
Abstract:
|
Planning successful research projects with both high power and minimum sample size is an important step in any study (Kraemer & Blasey, 2015). When working with longitudinal data, which are extensively used across disciplines to model changes over time, more advanced models as well as their respective power estimation tools are required to account for correlation among observations. When dealing with continuous longitudinal responses, many studies have been conducted using GLM; however, no extensive studies on discrete responses over time have been completed. These studies require more advanced models within random effects, conditional, and marginal models (Firzmaurice et al., 2009). Examples of these models include generalized linear mixed model, generalized estimating equation, and generalized method of moments. The purpose of the current study is to assess power analysis options for aforementioned models with varying types of responses using likelihood ratio, Wald, and score test statistics based on previous work conducted by Rochon (1988), Liu and Liang (1997), and Lyle et al. (2006). Options within R and SAS are recommended and used for comparing different power techniques.
|
