Sep 19, 2013
This entire episode highlights the dangers of focusing on the extreme values in any analysis. Generate 50 random numbers from any distribution, do a statistical test with a 90% confidence and likely you will be able to conclude that five those numbers are “statistically different” from the mean of the underlying distribution. The ones that are “statistically different” are the ones that will be highlighted in the media as doing far better or worst than most of the other states.
Understanding reported changes in median income » Good Stats Bad Stats