SIMULATION STUDY OF HIERARCHICAL REGRESSION

Hierarchical regression - which attempts to improve standard regressio n estimates by adding a second-stage 'prior' regression to an ordinary model - provides a practical approach to evaluating multiple exposure s. We present here a simulation study of logistic regression in which we compare hierarchical regression fitted by a two-stage procedure to ordinary maximum likelihood. The simulations were based on case-contro l data on diet and breast cancer, where the hierarchical model uses a second-stage regression to pull conventional dietary-item estimates to ward each other when they have similar levels of food constituents. Ou r results indicate that hierarchical modelling of continuous covariate s offers worthwhile improvement over ordinary maximum-likelihood, prov ided one does not underspecify the second-stage standard deviations.