Smallest effect size of interest
Also known as: Minimal important difference
The smallest effect big enough to actually care about — set before you look.
What it means
The smallest effect size of interest is the threshold magnitude that a researcher specifies in advance as the minimum effect worth detecting or worth treating as practically meaningful. Defining it converts the vague question of whether an effect 'exists' into the sharper question of whether it is at least as large as something that matters, which disciplines study design and interpretation in several ways. It is the basis for an a priori power analysis, since one powers a study to detect effects of that size or larger; it sets the equivalence bounds in tests that aim to show an effect is negligibly small; and it guards against being impressed by statistically significant but trivial effects in large samples. The threshold can be justified by practical or clinical relevance, by cost-benefit reasoning, or by what prior interventions have achieved, and choosing it is a substantive judgment rather than a statistical computation. It matters because pre-specifying what would count as a meaningful effect is central to credible, decision-relevant research and to interpreting null results.
Examples
Before running the test, a team decides a new checkout design must lift completed purchases by at least half a percentage point to justify the engineering. Anything smaller counts as nothing, however significant.
A regulator states in advance how many points on the pain scale a new painkiller must beat placebo by, because with a large enough trial a difference no patient could feel will still reach significance.
First described in Promoted in psychology by Daniël Lakens.