An Overview of Some Advances in Epidemiologic Methods in the Past Quarter-Century
• Sander Greenland
Professor, Department of Epidemiology
UCLA Public Health – Epidemiology, Los Angeles, CA, United States

Since the early 1980s, epidemiology has been enriched by methods previously unknown within the field or only known in fragments. This talk highlights several methodologies that have made inroads into teaching and practice.

Bias analysis has its roots in basic sensitivity formulas for confounding that were first developed by Cornfield and Bross for the tobacco controversy of the 1950s. In more recent decades it has borrowed from the risk-analysis literature and expanded into a field that lays bare defects in conventional statistical inference. In tandem, causal diagrams have emerged as a means for visually locating causal and selective forces that lead to bias in causal inference.

Along a separate line of descent emanating from general statistics, Bayesian and empirical-Bayesian methods have slowly penetrated epidemiology, often under the guise of hierarchical, multilevel, mixed, or random-coefficient models. These methods have helped to answer controversial questions about how to handle multiple exposures and confounders, and how properly incorporate contextual (prior) information into analyses.

Among developments originating in epidemiology, valid methods for causal analysis of time-varying exposures were all but unknown before Robins discovered the g-computation formula in the mid-1980s. During the ensuing decades, he and his colleagues have elaborated methods based on this formula into a full subspecialty of statistics.

Several valuable new study designs also emerged over the period. Chief among them, case-crossover designs have been embraced widely. Unfortunately, two-stage (two-phase) sampling methods have not yet penetrated as far. Some reasons for the difference will be suggested.

Sander Greenland is Professor of Epidemiology and Statistics, University of California, Los Angeles.  Professor Greenland has been a leading contributor to epidemiologic statistics, theory, and methods for over three decades.  His major research contributions include assessment of selection bias, misclassification, and confounding effects in epidemiologic research, and critical evaluation of statistical methods for observational studies.  He has authored or co-authored over 300 articles in the health sciences and statistics, as well as the textbook Modern Epidemiology, and has given invited presentations at universities and conferences throughout the world. He has served as an associate editor for several leading journals, and is a Fellow of the American Statistical Association and of the Royal Statistical Society. Professor Greenland received Bachelor's and Master's degrees in mathematics from the University of California Berkeley, and Master's and Doctoral degrees in Epidemiology from the University of California Los Angeles.