Hacker News Reader

Comment by mg | Hacker News Reader

There is a lot of confusion on how to interpret studies.

The reasoning of this article seems to be "The study did not produce a significant result, therefore the treatment is ineffective".

But that is not how to think about significance.

Otherwise, you could show for any treatment that it is ineffective. By simply doing a study small enough to produce an insignificant result.

View on HN

True, but in this case the study sponsor is financially motivated to find a non-null result, not a null result.

This mistake is so unbelievably frequent. Need a concise quip like “correlation isn’t causation”.

Yes, that might be helpful.

First attempt:

    Insignificant does not mean ineffective

Not as good as "correlation isn't causation" though. I wonder if there is a better quip. Or if the concept is too complex to wrap in a catchy statement.

“Insignificance isn’t ineffectiveness”

Reply on news.ycombinator.com