If you read the preprint that first blog links to, they didn’t actually find a negative efficacy with any kind of confidence. They find large confidence interva

If you read the preprint that first blog links to, they didn’t actually find a negative efficacy with any kind of confidence. They find large confidence intervals that include zero, with large(ish) p-values to boot. P-values and multiple comparison discussion aside, calling this evidence of negative efficacy is disingenuous at best. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3949410 https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3949410