The Design and Analysis of the Observational Study—A Review

In recent years the number of studies which attempt comparisons without adequate randomization has increased rapidly, particularly in disciplines and areas of research involving human populations. This article highlights some of the more important developments (and emerging needs) in the design and analysis of such investigations. It provides a historical overview of those techniques adapted to bias reduction, covers recent developments in contingency table analysis, particularly as this relates to the identification and estimation of possible bias components, and considers recent work on the statistical evaluation of methods of bias removal. Since it is clearly relevant to considerations of bias, pertinent work on measurement error is also included.