Matuschek, et al. (2017). Balancing Type I Error and Power in Linear Mixed ModelsHannes Matuschek, Reinhold Kliegl, Shravan Vasishth, Harald Baayen, Douglas Bates
Abstract: Linear mixed-effects models have increasingly replaced mixed-model analyses of variance for statistical inference in factorial psycholinguistic experiments. The advantages of LMMs over ANOVAs, however, come at a cost: Setting up an LMM is not as straightforward as running an ANOVA. One simple option, when numerically possible, is to fit the full variance-covariance structure of random effects (the maximal model; Barr et al., 2013), presumably to keep Type I error down to the nominal $\alpha$ in the presence of random effects. Although it is true that fitting a model with only random intercepts may lead to higher Type I error, fitting a maximal model also has a cost: it can lead to a significant loss of power. We demonstrate this with simulations and suggest that for typical psychological and psycholinguistic data, models with a random effect structure that is supported by the data have optimal Type I error and power properties.
Also available at arXiv.