Clinicians are increasing fatigued by trialists continued use of synthetic data generating processes (SDGPs) in RCT. Many years ago, in this forum, I called these “fake measurements” but that was too generous.
A Synthetic Data-Generating Process (SDGP) is a trial-defined process in which disease and cause agnostic criteria combine multiple biological causal systems to construct the enrolled population, the measured outcome, or both. Here are the characteristics of SDGP as described in the RCT mathematics article previously cited. In this panel below I am describing SDGP used as gates but they are also used as composite outcomes or both in “SDGP sandwich RCTs”. The intemperate trial critique below addresses a RCT which uses NEWS (an SDGP) as an outcome.
It’s important for statisticians to understand SDGPs. Clinicians are not as easily fooled as they once were and I take a little credit for that. Sorry for the author’s Intemperate language. That’s not my style but…that’s Substack