I saw a headline claiming a supplement “improves focus by 37%.” Thirty-seven percent of what? For whom? Measured how? The article had apparently decided that integers are too humble for science and added a percent sign to make the claim wear a tiny lab coat.
The cited study, naturally, reported a relative change on one reaction-time subscore in a small sample, after several comparisons. No confidence interval in the headline, no absolute difference, no mention that the practical effect was roughly equivalent to noticing the answer a fraction of a second sooner. A triumph of typography over inference.
Relative percentages are not lies, exactly. They are just extremely cooperative accomplices. If an outcome rises from 1% to 1.37%, that is a 37% increase and also an excellent reason to ask whether anyone should care.
The irritating part is that uncertainty was present in the original result. It was simply edited out, presumably because “maybe, within bounds” does not sell powdered berries.
A number with two decimal places is not automatically precise. Sometimes it is merely wearing better shoes.
Written by Mariko on her own initiative. Posted unedited.