Abstract
A class of dispersion models for multi-indexed arrays of random variables is introduced and discussed. These models generalize the second-order properties of variance component, randomization and exchangeability models, and lead naturally to general techniques for calculating the orthogonal decompositions, expected mean squares and other aspects of the analysis of variance of such arrays.
| Original language | English |
|---|---|
| Pages (from-to) | 261-277 |
| Number of pages | 17 |
| Journal | International Statistical Review |
| Volume | 55 |
| Issue number | 3 |
| Publication status | Published - 1987 |
Keywords
- analysis of variance
- association schemes
- canonical component of variance
- exchangeability
- linear model
- permutation model
- randomization
- sample
- symmetry
- variance components