Balancing Type I error and power in linear mixed models

摘要: Linear mixed-effects models have increasingly replaced mixed-model analyses of variance for statistical inference in factorial psycholinguistic experiments. Although LMMs many advantages over ANOVA, like ANOVAs, setting them up data analysis also requires some care. One simple option, when numerically possible, is to fit the full variance-covariance structure random effects (the maximal model; Barr et al. 2013), presumably keep Type I error down nominal alpha presence effects. it true that fitting a model with only intercepts may lead higher error, has cost: can significant loss power. We demonstrate this simulations and suggest typical psychological data, power achieved without inflating rate if selection criterion used select effect supported by data.