1) The document describes methods for detecting influential observations in generalized linear mixed models (GLMMs) with overdispersion. 2) Three approaches are discussed for deriving local influence diagnostics: a closed-form expression of the marginal likelihood, an integral-based approach, and fully numerical derivations. 3) The local influence diagnostics can be decomposed into interpretable components that provide insight into the influence of individual subjects. The methods are demonstrated on a Poisson GLMM analysis of an epilepsy clinical trial dataset.