Rofecoxibpaper And H

Rofecoxib Paper (Studies.. version)

I may need to take into account Howson and Urbach. In particular, perhaps I may need to outline what I mean by reassurance in the context of shifting likelihood and why this reassurance will not mean much to H&U.

H&U and Harvard Medical School case - H&U argue the case that probabilities and finitine frequencies need not be linked. While each adheres to the probability calculus, cross-over between the two is fallacious.

P 25 We can infer that even if e had been generated by an experiment in which e is predicted by h but very unlikely were h to be false, that would still by itself give use no warrant to conclude anything about the degree of certainty we are entitled to repose in h. (cf Mayo 1996 P 177)

H&U - seem to focus on the fact that there is no LOGICAL connection between significance levels and the truth or falsity of the hypothesis (rely on Harvard Medical School example)… is a looser connection possible? i.e. CS holds that low p-values say something (while at the same time may be just an unlikely result)… in holding that CS may say ‘something’ the shifting likelihood condition helps save CS in situations where it is the case that the significance test is more likely to lead us astray

Does experimentalism provide an avenue? Accept the fact that there is no logical connection between significance levels and the truth or falsity of the hypothesis under question. CS methods look to set up a test of a hypothesis. Given auxiliary hypotheses AND the experimental set-up (intervention) an experimental result can be predicted. Classical methods provide a way of thinking about the result which did actually occur and whether it is inline with the expected result. Can ALWAYS either highlight something about an underlying process or just be a chance result. CS methods and the experimental set up are such that when a result different from that expected is observed the possibility that it is a chance result is minimised.