Multilevel models

Multilevel models are a type of regression analysis that allows modeled relationships to be studied at multiple levels, simultaneously and to vary according to the context in which they are situated. They are used with data drawn from hierarchically structured populations to explore the presence or otherwise of contextual effects and to help explain any tendency for groups and sub-groups to exhibit similar characteristics and outcomes. In geographic applications, the levels of the model can represent differences of scale, fitting well with traditional geographic interest in scale dependent relationships, spatial variation, and the interactions between people and places. They can also be temporal, allowing for longitudinal analysis; or spatiotemporal, mixing the two. Multilevel modelling is distinctive in having an object-oriented view of geographic space and a hierarchical understanding of geographic relationships. This contrasts with field-based approaches and those that conceive spatial relationships in terms of their distances apart. However, recent developments have begun to combine the approaches. This chapter provides examples of hierarchical relationships and of some types of multilevel model, offers a rationale for using multilevel models instead of simpler regression, provides a case study with some demonstrative analysis, and discusses some of the advantages and disadvantages of adopting a multilevel approach in research.