Measurement Error, Statistical Power, and Design Efficiency in Planning Cluster-Randomized Trials

• In Event: Hilton Roundtable Session 1
  In Roundtable Session: Methodological Advances in Experimental Design

Abstract

Statistical power and design efficiency are two key considerations that researchers must address in planning experiments. Both considerations are integrated in an optimal design framework, which identifies the sample allocation that produces the maximum statistical power under a fixed budget. However, measurement error, which impacts both statistical power and design efficiency, has not been integrated in power analysis and design efficiency framework. In this study, we extend the previous optimal design framework by adjusting for multilevel reliability information in power analysis and optimal sampling framework for cluster-randomized trials. The proposed framework can help researchers accurately design cluster-randomized trials with enough statistical power and design efficiency. The R package can help researchers’ calculation.

Authors

• Zuchao Shen, University of Georgia

• Walter L. Leite, University of Florida

• Benjamin Kelcey, University of Cincinnati