I did an analysis of how convincing the Oxford-AstraZeneca claim of 90% effectiveness is.
Unfortunately I inferred the numbers of infections in each group incorrectly according to this—the infections were split 3:27 between the half-full group and the full-full group, not 2:28 as I’d calculated. (Note that the naive interpretation of the numbers doesn’t come to 90% or 62% effectiveness so I assume they’re doing some corrections or something else which alters the result slightly.)
That means the 8:1 Bayes factor I originally calculated (in favour of half-full being more effective vs the two different regimens being equally effective) comes down to 2.9:1. In my book that isn’t enough evidence to overcome the prior against the half-full dose regimen being more effective.
The above assumes that everything else about the groups is equal.
Having read the report linked in the OP I think the actual update should be noticeably lower, particularly as the half-full treatment group were younger than the full-full treatment group (or at least only the latter included anyone >55 years old).
Will the half-full dose regimen for the Oxford-AstraZeneca vaccine prove to be more effective than the full-full dose regimen?
(I mean which interpretation will the evidence favor, not on whether they go ahead with the half-full as the standard dose)
I did an analysis of how convincing the Oxford-AstraZeneca claim of 90% effectiveness is.
Unfortunately I inferred the numbers of infections in each group incorrectly according to this—the infections were split 3:27 between the half-full group and the full-full group, not 2:28 as I’d calculated. (Note that the naive interpretation of the numbers doesn’t come to 90% or 62% effectiveness so I assume they’re doing some corrections or something else which alters the result slightly.)
That means the 8:1 Bayes factor I originally calculated (in favour of half-full being more effective vs the two different regimens being equally effective) comes down to 2.9:1. In my book that isn’t enough evidence to overcome the prior against the half-full dose regimen being more effective.
The above assumes that everything else about the groups is equal.
Having read the report linked in the OP I think the actual update should be noticeably lower, particularly as the half-full treatment group were younger than the full-full treatment group (or at least only the latter included anyone >55 years old).
(I mean which interpretation will the evidence favor, not on whether they go ahead with the half-full as the standard dose)
There’s also a Metaculus question about this: