Hey all, I asked this over in the main maths room but didn't get any takers. Perhaps it's too simple a question. I'm just working out some sample size and power problems. I have the question and solution right here, but I just don't know where that 0.10 is coming from? "This leads to"... what leads to it, exactly? i.ibb.co/c1g0rXD/Q2-Question-and-solution.png
@jserv: that is a bona fide question for our main site. Please use the self-study tag and read (and follow) its info before posting. In particular, please type the relevant parts of your question, do not just upload a screenshot. Good luck!
@MartijnWeterings & @StephanKolassa Agreed, it's pretty depressing. I suppose the best that can be said is that it didn't have any real-world impact (to my knowledge). Unlike the ghastly Reinhart-Rogoff paper that shaped global economic policy for several years, and Brian Wansink's body of work that still drives tens of millions in nutrition spending based on completely bogus work.
@jserv: The value 0.10 relates to the confidence interval which is 4.9 to 5.09 or roughly the mean 4.99 mmol plus minus the value 0.10. However that answer made a small mistake. The 95% confidence interval is +- 2 times the SE. The answer used only 1 times the SE.
@mkt Not only was the R-R paper wrong on the substance (their conclusions were only correct because of their accumulated errors; doing the analysis without errors reverses their conclusion) but after their statistical claims collapsed, the authors insisted that they were still right but for other reasons.
"My argument is true because of statistical analysis but if the statistical analysis shows my argument is backwards then the argument is true for other reasons" is a strange way to think about evidence and truth claims.
@Sycorax I was aware of multiple errors (arbitrary & uneven binning of a continuous variable, Excel coding error) that each would have killed their stated result, but not that response from them. Astonishing chutzpah.
On the bright side, I have seen a large improvement in statistical practice in my field in the last ~5-8 years, especially among young scientists. Still rather short of satisfactory, but things have been moving in the right direction. A lot more awareness of the worst practices, even if things are still somewhat hamstrung by a continued conflation of statistical and scientific hypotheses (and the inferential & p-value mess that flows from that)
