Abstract:
|
Variance components for measurement systems analysis (MSA) models are typically estimated using either Expected Means Squares (EMS) or Restricted Maximum Likelihood (REML). Because negative variance component estimates do not make sense in MSA studies, these estimates are set to zero when these estimation methods are used. In addition, Portnoy and Sahai introduced a Bayesian method that produces only strictly positive variance components. We previously compared the ability of REML and the Bayesian method to estimate variance components for typical MSA models. In that study we found that both methods frequently had difficulty correctly classifying a measurement system. In this study, we investigate different sample sizes of measurement systems studies and identify the minimum sample sizes needed to correctly determine the adequacy of a measurement system with high probability.
|