Transitivity, Incoherence, and Indirect Comparisons
by Thomas Lumley
Abstract: Direct comparisons in randomized experiments are the gold standard for comparing interventions in medicine and other areas of science. The appropriate use of indirect comparisons, when two interventions have not been compared directly but each has been compared to a third, is controversial. I will discuss how misleading results in indirect comparisons can arise from deficiencies in the data or in the tests and estimators used. I will describe network meta-analysis as a technique for combining direct and indirect comparisons from clinical trials. I will also present some results on characterising transitive two-sample tests. If time permits, I will make some poorly-substantiated generalisations about the teaching of rank tests.
For More Information: Owen Jones tel. 8344-6412 email: email@example.com