A Bayesian Approach to Schools' Catchment-based Performance Modelling
Abstract:A Bayesian model for predicting schools' performance in SATs is proposed, in which geographical explanatory variables are averaged over a radius centred on each school. The coefficients for the explanatory variables, and the averaging (or smoothing) radius are treated as unknown variables, whose prior and posterior distributions will be considered in the Bayesian analysis. Posterior distributions cannot be determined analytically, and will be investigated using simulation via Gibbs sampling.
Document Type: Regular Paper
Publication date: 2001-05-01