The benefits of using semi-continuous and continuous models to analyze binge eating data: A Monte Carlo investigation

Change in binge eating is typically a primary outcome for interventions targeting individuals with eating pathology. A range of statistical models exist to handle these types of frequency distributions, but little empirical evidence exists to guide the appropriate choice of statistical model.

Monte Carlo simulations were used to investigate the utility of semi-continuous models relative to continuous models in various situations relevant to binge eating treatment studies.

Semi-continuous models yielded more accurate estimates of the population, while continuous models were higher powered when higher levels of missing data were present.

The present findings generally support the use of semi-continuous models applied to binge eating data, with total sample sizes of roughly 200 being adequately powered to detect moderate treatment effects. However, models with a significant amount of missing data yielded more favorable power estimates for continuous models.