8 ms·
One 'scientific villain' that occurs to me: Ronald Fisher. He used his considerable influence in the statistics community to suppress Bayesian inference researc
by arcanus 11y ago
One 'scientific villain' that occurs to me: Ronald Fisher. He used his considerable influence in the statistics community to suppress Bayesian inference research because he disagreed with the premise of prior probability. Set Bayesian research back decades. He was a strong proponent of the p-value causing some of the reproducibility crisis in scientific publications.
See: "The theory that would not die" by Sharon Bertsch for a very entertaining read of the entire conflict.
Full disclosure: I've published papers in Bayesian statistics.
- skj 11y agoWithout specifically disagreeing with you, the real advance that led to the bloom in Bayesian statistics was fast computers. MCMC and Gibb's sampling, and variational interference are what got the intractable-when-useful field of Bayesian statistics back on its feet.
- arcanus 11y agoI think you are certainly right that computation has really permitted Bayes to shine. But I think the stigma against the study persisted for decades: we had Metropolis-Hastings in the 70s, after all. And as the book argues out, Bayes, even when limited to non-sampling techniques such as conjugate priors, was still useful for a variety of interesting problems in cryptography, search theory, etc.