The law of large numbers defied
The law of large numbers says that a larger sample gives a better picture of the whole group. In other words, if you ask 1,000 Americans who they plan to vote for, you'll get a better idea than if you ask 100.

But if your sample is biased, as it was in the infamous Literary Digest poll of 1936, the increasing number in the sample only exaggerates the error.

As they say in stats, there's no error like a systematic error (defined). (They don't say that, I just invented it. But it sounds convincing, doesn't it?).

home back
NISE/NSF