Preface
Throughout a long career as a statistician, I have frequently found myself wrestling, in one way or another, with issues of bias and causation. As a methodologist, researcher, consultant, or expert witness, I have had to propose, justify, or criticize many varieties of causal statements. My training in mathematics and statistics prepared me well to deal with many aspects of the diverse, and occasionally bizarre, problems I have chanced to encounter. However, the statistical theory I studied in graduate school did not deal explicitly with the subject of causal inference, except within the narrow confines of randomized experimentation.
When I entered the “real world” of statistical research and consulting, the problems I regularly faced were not amenable to strict experimental control. They typically involved causal effects on human health and behavior in the presence of observational data subject to many possible sources of bias. To attack these problems, I needed analytic weapons that were not in my statistical arsenal. Little by little, I found myself being transformed into a practitioner of some dark art that involved statistics, but that drew as well on intuition, logic, and common sense.
The nature of this evolution can be best illustrated by an anecdote. The first legal case in which I provided statistical expertise was an employment discrimination lawsuit against a Boston-based Fortune 500 company. The plaintiffs were convinced that black workers were being systematically ...
