# It is conservative to assess significance by the degree of overlap of confidence intervals

February 25, 2013

~~Here~~ Here is a really interesting paper with a very simple message:

1. When two confidence intervals do not overlap, then the two parameters are significantly different.

2. When two confidence intervals overlap, then the two parameters may or may not be significantly different.

In other words, when there is no overlap, you don’t really need to test for significance. But when there is overlap, you do need to test for significance.

Hope I’m not oversimplifying.

Advertisements

2 Comments
leave one →

Unfortunately, your our eyeball is not a well-defined statistical procedure. It’s always best to use the appropriate hypothesis test instead. Overlapping confidence intervals do not mean two values are not significantly different. There’s more information here on that: http://www.statisticsdonewrong.com/significant-differences.html

“Overlapping confidence intervals do not mean two values are not significantly different.”

I know. That’s what I meant by “When two confidence intervals overlap, then the two parameters may or may not be significantly different.”