Statistics for Aggregate Variationist Analyses

This chapter presents methods for the analysis of large aggregates and techniques that seek to identify groups together with their common speech habits. Jean Seguy examined the distribution of aggregate differences (in Gascogne) as a function of geography, displaying a sub‐linear curve, which John Nerbonne argues contradicts Peter Trudgill's gravity theory of dialect divergence. These early works motivate the aggregate perspective. Regression analyses have played a role in characterizing the relation between geographic distance and aggregate linguistic differences. Another approach, generalized additive modeling (GAM), is able to simultaneously detect the aggregate geographical pattern, whereas also identifying the importance of other relevant social and lexical predictors. The advantage of the GAM is that it allows one to directly incorporate the complex influence of geography on the aggregate patterns, whereas simultaneously considering the importance of other social and lexical variables.