Power Analysis Is Too Damn Hard! Good-Enough Sample Size Determination for Structural Equation Modeling and Beyond

• In Event: Statistical Grab Bag

Abstract

Statistical power, while absolutely essential for inferential statistics, depends on many factors, posing great challenges for power analysis in practice. With conventional approaches, researchers need to meticulously set a population value for each model parameter—not only for the focal parameters for which testing is of primary interest, but also for each of the peripheral parameters that set the focal parameters’ context. To help mitigate these barriers, the current study proposes a simplified and flexible framework for a priori power analysis (i.e., sample size determination) that greatly reduces the burden of specifying a fine-grained model structure and the corresponding population parameter values, while preserving the validity of sample size/power estimates. The theoretical framework is discussed along with an illustrative example.

Authors

• Yi Feng, University of California - Los Angeles
  Presenting Author

• Gregory R. Hancock, University of Maryland
  Non-Presenting Author