Roland Schäfer. 2018 (expected). Generalised Linear Mixed Models. In: Stefan Gries & Magali Paquot. Practical Handbook of Corpus Linguistics. De Gruyter. [Click to download draft.]

Mixed effects modeling – alternatively called *hierarchical *or *multilevel modeling *is a straightforward extension of (generalized) linear modeling as discussed in the previous chapter. A common characterization of mixed-effects modeling is that it accounts for situations where observations are *clustered *or *come in groups*. In corpus linguistics, there could be clusters of observations defined by individual speakers, registers, genres, modes, lemmas, etc. Instead of estimating coeffcients for each level of such a grouping factor (so-called fi*xed effects*), in a mixed model they can be modeled as a normally distributed random variable (a so-called *random effect*) with predictions being made for each group. This chapter introduces readers to the situations where mixed effects modeling is useful or necessary. Thee proper specification of models is discussed, as well as some model diagnostics and ways of interpreting the output. Readers are assumed to be familiar with the concepts covered in the previous chapter on (Generalized) Linear Models.