Robustness of nMAX for Sample Size Planning Under Nonnormality

• In Event: Innovative Statistical Methods and Computational Tools for Complex Research Designs

Abstract

A priori power analysis for measured variable models requires the population values of all model parameters. The many contextual parameters, whose values are typically unknown, however, can have a dramatic influence on sample size planning to test the focal parameters. Recently, Hancock and Feng (in press) developed an approach to collapse all contextual parameters into a single metaparameter that is set conservatively to yield a sample size nmax, which was derived analytically to ensure adequate power under the assumption of conditional normality. The current work aims to investigate the integrity of nmax power estimates to varying degrees of nonnormality, evaluating under what conditions it is robust and, if necessary, offering distributional corrections to use a priori in sample size planning.

Authors

• Gregory R. Hancock, University of Maryland

• Yi Feng, University of California - Los Angeles