Bayesian modeling for linguistic researchers This workshop will provide an introduction to Bayesian modelling for researchers in linguistics, including an outline of what makes a model “Bayesian”, some advantages of Bayesian modeling, and how to interpret Bayesian models reported in published articles. The intended audience for this workshop is researchers who are comfortable with multilevel (mixed-effects) models (e.g., lme4 in R), but may not be familiar with Bayesian approaches to these models. We will begin by contrasting Bayesian and frequentist estimation methods, and Bayesian inference to traditional methods based on null-hypothesis significance testing (e.g., p-values). We will then discuss the advantages to Bayesian approaches for the analysis of linguistic data, including the ability to build models with any number of “random” effects or slopes and the ability to easily build multilevel models for binomial, multinomial, and ordinal dependent variables. A Bayesian analysis (using JAGS in R) will be compared to an equivalent frequentist mixed-effects analysis using lme4, including a comparison of the outputs of these models and of the sorts of questions these models can (and can’t) help us answer. The code and data used for this analysis will be made available before the workshop so that participants can follow along if desired. Finally, we will outline some useful resources for those interested in learning to apply Bayesian models to their own data.
