Paper argues clinical trials should use minimax regret, not p-values, to guide treatment decisions
RexDouglass · x · 2026-07-29
- A paper summarized in the image argues that clinical research is often organized around statistical conventions rather than real treatment decisions.
- It criticizes hypothesis testing, confidence intervals, and common prediction metrics for being weakly connected to patient welfare, and says treatment evaluation should jointly consider benefits, harms, and patient preferences.
- The proposed alternative is statistical decision theory, especially a minimax-regret criterion, which weights both the probability of an error and the welfare loss from making it.
- The example claims that a melanoma trial could have required far fewer patients under an explicit welfare/regret formulation than under conventional power calculations.
More from AGI Musings
- AI wrote 40% of a 46-page business draft, but legal compliance still needed humans — AryHHAry · 2026-07-29
- Anti-pause stance: slowing AI progress is “anti-human,” says one reply — mark_k · 2026-07-29
- Bayesian intuition may reappear at scale as compute keeps growing — gleech · 2026-07-29
- AI safety post says the singularity is coming faster and alignment work is urgent — ericelliott_ · 2026-07-29
- A thread argues models may lose contact with humans when training is dominated by other models — zetalyrae · 2026-07-29
- Frontier post-training is a humanities problem as much as a technical one — brianryhuang · 2026-07-29