Two confidence interval approaches on the dependability coefficient in a two-factor crossed design.

For decisions based on the absolute level of performance among individuals, the dependability coefficient-a ratio of variance components-is used as a measure of reliability. Two methods developed for a two-factor random effects crossed (or one-facet) design-the Arteaga, Jeyaratnam, and Graybill (AJG) approach and the Cappelleri and Ting (CT) approach-are applied to construct confidence intervals on the dependability coefficient. A simulation study is conducted to investigate and compare the confidence interval coverage on the dependability coefficient based on AJG and CT. Both methods generally meet at least the nominal coverage. Both methods are illustrated with examples.