take U.S. elections. "polls indicate that so and so is ahead by 67%."
reaction, more people vote for him because if he is already ahead, he must be the right one, and he is going to win anyway, so i may as well vote for him.
they dont take into account that the poll was done in the candidates home state, and only 1000 people were polled. hmmmmm.....
the next week the guy wins by a land slide and everyone wonders why.
The Industriallist wrote:Statistics are reliable and accurate when you have large, representative sampling sets.
The problem is, you have to know what the statistics are actually telling you, and not claim more.
A lot of the bad use of statistics comes from either using a statistic for shock value without reference points or from looking for a single correlation and then acting as if that proves a claim about causation (The Hunter's earlier statistic on crime and race). You can possibly disprove causation with statistics, but you can't prove it.
).