Centre for History and Philosophy of Science seminar

Comparability, randomisation and causal inference

It is commonly held that causal inference in clinical research requires comparable groups and that the main purpose of random allocation is to achieve comparability. I argue that both claims are false. There are two distinct methods for inferring causation, one based on the method of difference and the other on statistical inference. Only the former requires comparability. The latter takes into account the probability that incomparable groups are generated and relies on random allocation in order to ensure that the treatment condition is the independent variable targeted by the allocation intervention.