Masters candidate in Biostatistics, Zachary Brown, will present:
“Efficient and Powerful Test for the Mean Parameter in the One-way Random-effects ANOVA Model with Application to Medical Device Comparison Studies”
Plan B Adviser: Baolin Wu
Abstract: In medical comparison studies, paired repeated measures are commonly used to evaluate the effectiveness and equivalency of some investigational device compared to a standard system. For analyzing such paired repeated measures, the one-way random effects model is generally adopted for estimation and inference. The estimation of such model has been well studied, however, the statistical inference has been a long standing problem. Closed-form analytical tests and confidence intervals do not exist for the mean parameter in an unbalanced study design. Instead, most existing methods have been based on approximate solutions. For example, a geometric mean-based approximation test is easy to compute but is non-efficient for unbalanced data. Bootstrapping is well regarded, but can be difficult to implement for studies with a large number of observations and lead to suboptimal performance. We build on the novel framework of generalized significance test and generalized confidence interval to introduce an efficient and powerful approach. We evaluate our new approach against the bootstrapping, and the geometric-mean approximation through simulation studies under different scenarios. We further demonstrate the utility of our proposed methods through application to a pulse oximetry comparison study and demonstrate the effectiveness of the new oximetry compared to the standard system.