NPR On The Media, May 13, 2011
Surprising and exciting scientific findings capture our attention and captivate the press. But what if, at some point after a finding has been soundly established, it starts to disappear? In a special collaboration with Radiolab we look at the 'decline effect' when more data tells us less about scientific truth. (17m 34s)
Listen to the podcast at
For a transcript, see www.onthemedia.org/transcripts/2011/05/13/04
See also this Wiki on the decline effect
Here is a great article in New Yorker on Decline Effect
It looks like a tip of the iceberg that can crash much of recent "soft science" analysis, based on small samples and many variables, typical in psychology, medicine, social studies ...
The "Decline effect' seems to be a clear case of overfitting in the original data analysis