IGI Global, Pennsylvania, 2023
Ceiling and floor effects are frequently encountered in clinical studies that measure factors such as psychology, quality of life, or postsurgical well-being. In the literature review, in most of the studies, the authors observed that these observations with ceiling and floor effects are ignored and classical statistical methods are applied. However, the results obtained do not reflect the real data but cause higher standard errors. The first aim of this article is to fit a regression model by imputing missing values to the censored observations, instead of removing them from the data in the estimation of the outcome with ceiling and floor effects. The authors also aim to compare Tobit regression, zero-inflated Poisson regression, ordinary least-square regression, and the suggested regression-based imputation method. In the application, the performances of two data sets were verified by cross-validation. Findings from both ceiling and floor effective outcomes have shown that the regression-based assignment method provides the best performance.