The “numerical method ” in medicine goes back to Pierre Louis ’ study of pneumonia (1835), and John Snow’s book on the epidemiology of cholera (1855). Snow took advantage of natural experiments and used convergent lines of evidence to demonstrate that cholera is a waterborne infectious disease. More recently, investigators in the social and life sciences have used statistical models and significance tests to deduce cause-and-effect relationships from patterns of association; an early example is Yule’s study on the causes of poverty (1899). In my view, this modeling enterprise has not been successful. Investigators tend to neglect the difficulties in establishing causal relations, and the mathematical complexities obscure rather than clarify the assumptions on which the analysis is based. Formal statistical inference is, by its nature, conditional. If maintained hypotheses A, B, C,... hold, then H can be tested against the data. However, if A, B, C,... remain in doubt, so must inferences about H. Careful scrutiny of maintained hypotheses should therefore be a critical part of empirical work—a principle honored more often in the breach than the observance. Snow’s work on cholera will be contrasted with modern studies that depend on statistical models and tests of significance. The examples may help to clarify the limits of current statistical techniques for making causal inferences from patterns of association. 1.

This paper (which is mainly expository) sets up graphical models for causation, having a bit less than the usual complement of hypothetical counterfactuals. Assuming the invariance of error distributions may be essential for causal inference, but the errors themselves need not be invariant. Graphs can be interpreted using conditional distributions, so that we can better address connections between the mathematical framework and causality in the world. The identification problem is posed in terms of conditionals. As will be seen, causal relationships cannot be inferred from a data set by running regressions unless there is substantial prior knowledge about the mechanisms that generated the data. There are few successful applications of graphical models, mainly because few causal pathways can be excluded on a priori grounds. The invariance conditions themselves remain to be assessed.

Did you know that the US has a higher death rate than Mexico? It’s a fact. In 2003, the death rate was 80 % higher in the US than in Mexico (8.4 per 100,000 versus 4.7). What does this statistic mean? Does Mexico have better health care than the US? That seems very unlikely. Yet it is difficult to claim that this unexpected relationship is due to chance, error or bias. The populations being studied are large; death is definite and usually counted accurately. You may be perplexed, confused or confounded when you learn that death rates are even lower in Ecuador and Saudi Arabia (4.3 and 2.7). An alternate explanation is confounding. Last (1995) defines confounding as “a situation in which the effects of two processes are not separated.” Confounding reflects the influence of a lurking variable. A lurking variable is often referred to as a confounder which Last defines as “a variable that can cause or prevent the outcome of interest … and is associated with the factor under investigation.” In comparing these death rates, a lurking variable may be the difference in the age distributions. Mexico has a much younger population than the US. In 2003, people under 15 were 50% more prevalent in Mexico than in the US (32% compared to 21%). People 65 and older were more than twice as prevalent in the US as in Mexico (12% compared to 5%). It’s a fact that older people are much more likely to die than younger people. Unless we take age distribution into account, a comparison of these crude (unadjusted) death rates may be misleading. Mexico’s comparatively low death rate is most likely due to its youthful population, rather than to its health care system. So how can we untangle this confusion? How can we “take into account’ the influence of a lurking variable that confounds an association?

Experiments offer more reliable evidence on causation than observational studies, which is not to gainsay the contribution to knowledge from observation. Experiments should be analyzed as experiments, not as observational studies. A simple comparison of rates might be just the right tool, with little value added by “sophisticated” models. This article discusses current models for causation, as applied to experimental and observational data. The intention-to-treat principle and the effect of treatment on the treated will also be discussed. Flaws in per-protocol and treatment-received estimates will be demonstrated.

Regression models are often used to infer causation from association. For instance, Yule [79] showed – or tried to show – that welfare was a cause of poverty. Path models and structural equation models are later

This paper presents a qualitative case study of statistical practice in a universitybased statistical consulting centre. Naturally occurring conversations and activities in the consulting sessions provided opportunities to observe questions, problems, and decisions related to selecting, using, and reporting statistics and statistical techniques in research. The consulting sessions provided simultaneous opportunities for consultants and clients to learn about using statistics in research. Consistent with contemporary theories that emphasize social dimensions of learning, major themes relate to (a) types of clients and consulting interactions, (b) disciplinary and statistical expertise, and (c) the role of material objects and representations. Evidence shows that consultants and clients learned during the consulting sessions and that the statistical consulting centre contributed positively to teaching and research at the university.

19,000 incl. abstract, references & notes * I have profited from the rancorous discussions of the Highland Park Colloquium on Theory, Methods, and Beer. I would also like to thank Neil Fligstein, Matt George, Ann Mische, and the reviewers for their probing criticisms that greatly increased the coherence of the argument, though all called for a more complete theoretical specification than I was able to provide. Finally, one can only acknowledge the loss of Pierre Bourdieu—it seems impossible to adequately describe how great a loss this is for the social sciences. What is Field Theory? Field theory is a more or less coherent approach in the social sciences, although the main directions of field theory have not been systematically integrated. The essence of field theory in the social sciences is the explanation of regularities in individual action by recourse to position vis-à-vis others. Position in the field is in turn considered to indicate the potential for a force exerted on the person, but a force that impinges “from the inside ” through motivation as opposed to through external compulsion. Motivation is accordingly considered to be the paramount example of social structure in action, as opposed to a residue of chance or freedom. Field theory

One of the most persistent research findings in criminology is that men commit much more crime than women. This typically is referred to as the gender gap in offending. Many researchers have noted that during the past several decades, women and men have converged in their rates of crime and the gender gap in offending has narrowed. Several explanations of this convergence have been proposed, including the economic marginalization hypothesis, which argues that the gender gap in crime has narrowed because women have experienced increasing economic hardship relative to men. This article reviews research on changes over time in the relative crime of women and men. It presents an analysis of Uniform Crime Reports data on the gender gap in offending from 1960 to 1997 and concludes that there has been an appreciable narrowing of the gap over this period in both property and violent offenses. The article then assesses the evidence in the criminological literature regarding the possible reasons for these changes, including the economic marginalization hypothesis. It then reviews evidence from demographic and economic research regarding the increasing marginalization of women and concludes that changes in the gender gap in crime are consistent with the findings of this research. Finally, the article proposes avenues for extending and refining the economic marginalization perspective. A B

The authors gratefully acknowledge support from the National Institute of Alcohol Abuse and Alcoholism. Any opinions expressed here are solely those of the authors and not of any sponsoring This study explores transitions between homeless and domiciled states. It describes the timing of departures from and returns to homelessness, and it tests theoretical propositions linking individual attributes and experiences to these transitions. Four theoretical frameworks guide the analyses: institutional disaffiliation, human capital deficiencies, personal disabilities, and acculturation. The data come from a longitudinal study of homeless individuals in Minneapolis. Various individual attributes are linked with leaving homelessness, including recent employment, welfare receipt, job training, identification with other homeless people, and homeless history. Fewer attributes are linked with returns to homelessness: work history and gender. These findings provide some evidence for existing explanations for homeless transitions, and they suggest promising avenues for further research on the dynamics of homelessness. The Dynamics of Homelessness Empirical studies of homeless people date back at least to the nineteenth century, and may now number well into the hundreds. These studies have served the important purpose of informing