
"No Difference" May Be the Wrong Conclusion, Scientists Warn
Image Credit: Courtesy of Universities of Manchester
Scientific Frontline: Extended "At a Glance" Summary: Equivalence Testing in Scientific Research
The Core Concept: Equivalence testing is a statistical approach that asks whether any observed difference in research data is too small to be meaningful, rather than simply asking if there is evidence of a difference.
Key Distinction/Mechanism: Instead of traditional statistical testing where a "non-significant" result (like a p-value > 0.05) is frequently misinterpreted as proof of "no effect," equivalence testing differentiates between effects that are truly negligible and results that are inconclusive due to insufficient data.
Origin/History: Highlighted in an August 2026 publication in PNAS by researchers from the Universities of Manchester, Oxford, and Arkansas, promoting the two one-sided tests (TOST) procedure.
Major Frameworks/Components:
- P-value misinterpretation: The common error of assuming a p-value greater than 0.05 means "no effect" rather than "insufficient evidence."
- Two one-sided tests (TOST): A specific equivalence testing procedure that is currently used in psychology and medicine but underused in other life and natural sciences.
- Practical equivalence: The requirement for researchers to define, before data collection, how small an effect must be to be considered scientifically or clinically uninteresting.

.jpg)




.png)






