Interpreting Blocks and Random Factors
Myra L. Samuels, George Casella, George P. McCabe
Journal of the American Statistical Association · 1991 · Volume 86 · Issue 415
DOI: 10.1080/01621459.1991.10475112
Abstract
Can blocks be tested in a randomized blocks design? It is well known that two different formulations of the linear mixed model yield conflicting answers to this question. This article examines the model formulations from the point of view of statistical relevance. Viewing blocking as a device to increase efficiency leads to a hypothesis concerning blocks that is distinct from the hypothesis of no “block main effect.” Relevant tests and interval estimates are reviewed. The merits of the two model formulations are compared, and recommendations are made to enhance the clarity and heuristic usefulness of mixed linear models.
Enable JavaScript for citation tools and additional paper interactions.