|
Abstract:
|
For mixed linear models having two variance components, inferences concerning the ratio of variances are often of primary interest. The variance ratio is related to the proportion of the overall variance due to a source of variation, and the proportion serves as a parameter which measures the importance of the source. Using normality assumptions, computing exact confidence intervals for the parameter under study is a straightforward process. However, difficulties arise when trying to choose an exact interval from a large collection of exact intervals. To select a particular confidence interval procedure, the author considers a quantity that is related to both the expected length and bias of an interval. Applications related to animal breeding studies are presented.
|
