ABSTRACT
Stratification is performed for two completely different reasons: to evaluate and possibly control for confounding by the stratification variable or, to analyze effect modification by the stratification variable. Confounding is a systematic error which may or may not be present in a particular study, while effect modification is connected with how different factors interact to cause disease. This chapter describes how stratification can be used to control for confounding. It looks at the instances both with and without effect modification. The chapter discusses the analysis of synergism and effect modification. The maximum likelihood method, or ML-method, is a general statistical method for estimation. Pooling is a weighing together of the stratum-specific estimates which approximately minimizes the variance of the estimator; the weights are chosen in proportion to the inverted value of the variance for each stratum specific estimate. The principles for standardization are the same regardless of the measure of disease occurrence used.
