:: Draft

1 point by light_hue_1 6 months ago | parent | context | next [–] | on: Psychological and psychiatric terms to avoid

(19) No difference between groups. … Authors are instead advised to write “no significant difference between groups” or “no significant correlation between variables.” This is terrible advice. To the public, and often even to experts, “significant” doesn’t mean “statistically significant” it means “big”. We need to abolish this use of “significant” not promote it. Way too many papers show “significant” (statistically significant) results that are not significant (so minor as to be irrelevant). This is the #1 source of misleading headlines.

dredmorbius 6 months ago | parent | next [–]

There’s been strong pushback against the term “statistical significance”, though I don’t seem to find a widely-accepted alternative. See e.g., “Moving to a World Beyond “p < 0.05”” Ronald L. Wasserstein, Allen L. Schirm, & Nicole A. Lazar. Pages 1-19 | Published online: 20 Mar 2019

https://www.tandfonline.com/doi/full/10.1080/00031305.2019.1…

“Statistically measurable” or “statistically determinable” come to mind though I can’t find a cite.

The originally intended … significance … of the term was that a difference was capable of being shown using statistical methods. Not that its size or context was itself significant in some semantic, practical, or other sense.

xmddmx 6 months ago | parent | prev | next [–]

You are correct - this is terrible advice. The main issue is that the word “significance” is overloaded with two meanings in research & stats:

“Large and important” (e.g. “clinical significance”, often given as an effect size, or some benefit/cost tradeoff)

“P value below alpha” (e.g “statistical significance”, Probabilty of rejecting H0, which is roughly equivalent to “if the true effect size were zero, what’s the chance I could see this effect size in my data given random fluctuactions [and a bunch of other assumptions]”