@trumanfrancis Yes. Thanks. Using the same controls more than once is not allowed unless you take account of the induced covariance between contrasts that are now not independent . That is to say, standard meta-analysis approaches will be wrong. https://t.
So far, Altmetric has seen
220 X posts from 135 X users, with an upper bound of 1,298,689 followers.
12,429 followers
12,429 followers
@AdanBecerraPhD @5_utr @f2harrell It was Don Rubin, I think, who drew attention to the implicit assumption in many RCTs that there are no different versions of the treatment. In meta-analysis it has relevance to the hypothesis one can test. See https://t.c
12,429 followers
@VickersBiostats @soboleffspaces I think that this is true of public sector work. I think pharma based statisticians are less keen on doing this. It is clearly legitimate for some purposes but not for all purposes. https://t.co/hZqEqaw1IZ https://t.co/a5l4
260 followers
RT @stephensenn: @adamkvonende @drjohnm @f2harrell @djc795 @statsepi As regards heterogenous pooling that is fine, provided that the null h…