Background: Multilevel modelling is a statistical technique that extends ordinary regression analysis to the situation where the data are hierarchical. Such data form an increasingly common evidence base for public health policy, and as such it is important that policy makers should be aware of this methodology.
Method: This paper therefore lays out the a basic description of multilevel modelling, discusses the problems of alternative approaches, and details the relevance for public health policy before describing which levels are relevant and illustrating the different kinds of hypotheses that can be tested using multilevel modelling. A series of examples is used throughout the paper. These relate to regional variations in the incidence of heart disease, the allocation of health resources, the relationship between neighbourhood disorder and mental health, the demand-control model in occupational health, and a school intervention to prevent cardiovascular disease.