In this article, the second article of the RCT structure analysis series, present the two species of RCT which have evolved and are now in common use. These two species of RCT are profoundly different and should not be conflated. The second species studies “synthetic data generating processes” (SDGPs) or their equivalents at a third layer estimand.
The three articles together describe the present spectrum of divergent RCT structures. Below a link is the second article in the RCT structure trilogy.
The third RCT structural analysis article in this series, linked here, is still a preprint.
This third article above, is the most important for clinicians and statisticians to understand because it discloses the discovery of the three estimand layers of many modern RCTs. One goal of this work is to bring clinicians and statisticians together discussing the origin and structure of the gate and the associated causal coherence of the trial as this has a major impact on the safety of the transportability of the gate triggered treatment into clinical space.
The catastrophe of the transport of ARDS meta analysis to generate “strong” ventilator guidelines of the COVID pandemic has been the catalyst for my efforts here to explain (after the smoke and pain has mitigated) why, mathematically, that happened, but it is up to those with the lectern, who teach medical statistics and trial theory to trialists and clinicians, to assure it does not occur again. Of course I’m always supportive of open debate of these issues. On the other hand anyone who thought this would all go away was wrong. Once you see the third layer estimand and the “cause mixture paradox”, you can’t unsee it.
The important discovery to teach is that one species of trials operates at a third layer estimand where unsafe transport is a real risk and must be addressed by design because the trial is anchored to a synthetic data-generating process (SDGP) rather than a causal biological system.
SDGPs are not what we are treating at the bedside.