Evidence-Based Medicine is a relatively new movement that seeks to put clinical med- icine on a firmer scientific footing. I take it as uncontroversial that medical practice should be based on best evidence-the interesting questions concern the details. This paper tries to move towards a coherent and unified account of best evidence in medicine, by exploring in particular the EBM position on RCTs (randomized controlled trials).

The evidence from randomized controlled trials (RCTs) is widely regarded as supplying the ‘gold standard’ in medicine—we may sometimes have to settle for other forms of evidence, but this is always epistemically second-best. But how well justified is the epistemic claim about the superiority of RCTs? This paper adds to my earlier (predominantly negative) analyses of the claims produced in favour of the idea that randomization plays a uniquely privileged epistemic role, by closely inspecting three related arguments from leading contributors (...) to the burgeoning field of probabilistic causality—Papineau, Cartwright and Pearl. It concludes that none of these further arguments supplies any practical reason for thinking of randomization as having unique epistemic power. IntroductionWhy the issue is of great practical importance—the ECMO casePapineau on the ‘virtues of randomization’Cartwright on causality and the ‘ideal’ randomized experimentPearl on randomization, nets and causesConclusion. (shrink)

In clinical and agricultural trials, there is the danger that an experimental outcome appears to arise from the causal process or treatment one is interested in when, in reality, it was produced by some extraneous variation in the experimental conditions. The remedy prescribed by classical statisticians involves the procedure of randomization, whose effectiveness and appropriateness is criticized. An alternative, Bayesian analysis of experimental design, is shown, on the other hand, to provide a coherent and intuitively satisfactory solution to the problem.

This paper considers a key point of contention between classical and Bayesian statistics that is brought to the fore when examining so-called ‘persistent experimenters’—the issue of stopping rules, or more accurately, outcome spaces, and their influence on statistical analysis. First, a working definition of classical and Bayesian statistical tests is given, which makes clear that (1) once an experimental outcome is recorded, other possible outcomes matter only for classical inference, and (2) full outcome spaces are nevertheless relevant to both the (...) classical and Bayesian approaches, when it comes to planning/choosing a test. The latter point is shown to have important repercussions. Here we argue that it undermines what Bayesians may admit to be a compelling argument against their approach—the Bayesian indifference to persistent experimenters and their optional stopping rules. We acknowledge the prima facie appeal of the pro-classical ‘optional stopping intuition’, even for those who ordinarily have Bayesian sympathies. The final section of the paper, however, provides three error theories that may assist a Bayesian in explaining away the apparent anomaly in their reasoning. (shrink)

“Absence of evidence isn’t evidence of absence” is a slogan that is popular among scientists and nonscientists alike. This article assesses its truth by using a probabilistic tool, the Law of Likelihood. Qualitative questions (“Is E evidence about H ?”) and quantitative questions (“How much evidence does E provide about H ?”) are both considered. The article discusses the example of fossil intermediates. If finding a fossil that is phenotypically intermediate between two extant species provides evidence that those species have (...) a common ancestor, does failing to find such a fossil constitute evidence that there was no common ancestor? Or should the failure merely be chalked up to the imperfection of the fossil record? The transitivity of the evidence relation in simple causal chains provides a broader context, which leads to discussion of the fine-tuning argument, the anthropic principle, and observation selection effects. (shrink)

A number of arguments have shown that randomization is not essential in experimental design. Scientific conclusions can be drawn on data from experimental designs that do not involve randomization. John Worrall has recently taken proponents of randomized studies to task for suggesting otherwise. In doing so, however, Worrall makes an additional claim: randomized interventional studies are epistemologically equivalent to observational studies, providing the experimental groups are comparable according to background knowledge. I argue against this claim. In the context of testing (...) the efficacy of drug therapies, well-designed interventional studies are epistemologically superior to well-designed observational studies because they have the capacity to avoid a type of selection bias. Although arguments for interventional studies are present in the medical literature, these arguments are too often presented as an argument for randomization. Randomization in interventional studies is defended on Bayesian grounds. (shrink)

Scientific reasoning is—and ought to be—conducted in accordance with the axioms of probability. This Bayesian view—so called because of the central role it accords to a theorem first proved by Thomas Bayes in the late eighteenth ...

It is sometimes thought that randomized study group allocation is uniquely proficient at producing comparison groups that are evenly balanced for all confounding causes. Philosophers have argued that in real randomized controlled trials this balance assumption typically fails. But is the balance assumption an important ideal? I run a thought experiment, the CONFOUND study, to answer this question. I then suggest a new account of causal inference in ideal and real comparative group studies that helps clarify the roles of confounding (...) variables and randomization. 1Confounders and Causes2The Balance Assumption3The CONFOUND Study 3.1CONFOUND 13.2CONFOUND 24Disjunction C and the Ideal Study 4.1The ultimate ‘other cause’: C4.2The ideal comparative group study4.3Required conditions for causal inference5Confounders as Causes, Confounders as Correlates6Summary. (shrink)

This volume of essays is very much a sequel to the two earlier collections by Jon Elster, Ulysses and the Sirens and Sour Grapes. His topic is rationality - its scope, its limitations, and its failures. Elster considers rational responses to the insufficiency of reason itself, and to the 'indeterminacies' in deploying rational-choice theory and discusses the irrationality of not seeing when, where, and what these are. A key essay which gives the collection its title examines disputes in cases of (...) child custody, which are paradigmatically indeterminate. Leaving aside cases where one parent is patently unfit and assuming that protracted dispute is against the immediate interests of the child, Elster argues that three options present themselves: a strong presumption in favour of the mother, a strong presumption in favour of the primary caretaker (also likely to be the mother), and tossing a coin. Though the first two options may be preferable in the short/medium term, Elster argues that there is a case for randomisation in the long term. The book will be read with interest by anyone concerned with political and social science. (shrink)

Medical nihilism is the view that we should have little confidence in the effectiveness of medical interventions. Jacob Stegenga argues persuasively that this is how we should see modern medicine, and suggests that medical research must be modified, clinical practice should be less aggressive, and regulatory standards should be enhanced.

Three main lines of arguments are presented as a defense of randomization in experimental design. The first concerns the computational advantages of randomizing when a well-defined underlying theoretical model is not available, as is often the case in much experimentation in the medical and social sciences. The high desirability, even for the most dedicated Bayesians, of physical randomization in some special cases is stressed. The second line of argument concerns communication of methodology and results, especially in terms of concerns about (...) bias. The third line of argument concerns the use of randomization to guarantee causal inferences, whether the inference consists of the identification of a prima facie or a genuine cause. In addition, the relation of randomization to measures of complexity and the possibility of accepting only random procedures that produce complex results are discussed. (shrink)