The Limits of Marginality

Abstract The “marginality principle” for linear regression models states that when a higher order term is included, its constituent terms must also be included. target article relies on this principle the fixed-effects part of mixed ANOVA designs and considers implication if extended to combined fixed-and-random-effects models, model selection tests specific some are not possible. We review basis delineate limits. then consider extension models. conclude we have been unable find in literature, including article, ourselves construct any satisfactory argument against use incomplete only could requires one assume it possible test point-null hypotheses, something disagree with, which believe incompatible with Bayesian model-selection methods article.