A Unified Approach to Simple Confidence Intervals for a Dozen Parameters Arising from the One-Way Random Effects Model

• Activity Number: 610
• Type: Topic Contributed
• Date/Time: Thursday, August 5, 2010 : 8:30 AM to 10:20 AM
• Sponsor: WNAR
• Author(s): Guangyong Zou*+ and Allan Donner and Julia Taleban
• Companies: The University of Western Ontario and The University of Western Ontario and The University of Western Ontario
• Keywords: delta method ; Wald-type ; limits of agreement ; generalized confidence interval ; variance ; lognormal

Abstract

The one-way random effects model is often adopted to analyze correlated continuous data, due to the fact that many parameters of interest are functions of the three parameters in the model. A dozen such parameters can be identified from the literature cutting across several disciplines in health sciences. Confidence interval estimation for these parameters has usually been perceived as difficult and thus tackled using computational intensive approaches. As an alternative, we present a unified approach to this problem based on recovering variance estimates from confidence limits for the mean and the two variance components. Besides being closed-form, an advantage of this approach is that it provides confidence intervals that reflect the shape of underlying sampling distributions. Simulation results show that this approach performs well in terms of overall coverage and tail errors.