The Influence of Number of Clusters and Level 2 Residual Distribution on Multilevel Parameter Estimates

• In Event: 31.038 - Power Analysis in Multilevel Modeling

Abstract

The number of clusters needed to ensure accurate parameter estimates in two-level models is a critical question because of the effort expended to obtain clusters in many studies and the cost of doing so. Despite research studying the impact of number of clusters (J) on estimation the values of J needed to provide accurate results in two-level models for normally-distributed cross-sectional data remains unclear. This situation becomes even less clear when level 2 residuals deviate from a normal distribution. This study employed Monte Carlo methods to explore the impact of systematically increasing J for several level 2 residual distribution and a two-level model. The results update and expand existing guidelines for the minimum J needed for accurate multilevel parameter estimates.

Authors

• Hao Jia, University of Minnesota

• Yadira Peralta, Centro de Investigación y Docencia Económicas

• Michael R. Harwell, University of Minnesota