Apr 12, 2007 16:15
An article in the NYTimes I was reading today made a startlingly useless claim. Here it is in unidentifiable terms:
The N effect is quite substantial. Some 15 percent of X can attribute Y, based on the assumption that 1 percent to 4 percent of X are Y, and that each additional N increases the odds of Y by 33 percent.
From my reading, the claim appears to be unsubstantiated but I guess it makes for great reading. Plus it is lots of fun to assume every piece of data to make a point.
arguments