Statistical Ties, and Why You Shouldn't Wear One

Andrew J. Vickers, PhD


January 18, 2007

In This Article

To the Polls

If I were to say the word "statistician" followed by the word "tie," you'd probably think of something brown and nylon. Or maybe grey. Either way, this is probably not the meaning intended when opinion pollsters tell you that 2 candidates are "in a statistical tie." As it happens, the concept of a "statistical tie" is an interesting way of looking at confidence intervals, a common and important concept in statistics.

So, having nothing better to do, one morning last fall I went out and asked 1000 people whom they would vote for in the forthcoming election for the Senate: 483 respondents favored the Republican candidate and 516 favored the Democrat (one guy said that he would be voting for a Klingon on the basis of intergalactic messages picked up through metal fillings in his teeth). Let's say that I had lunch and polled another 1000 in the afternoon. Now we wouldn't expect to get precisely the same results as we did on the first poll, indeed, if I found that exactly 483 of my afternoon sample were choosing Republican I'd probably be somewhat surprised. On the other hand, I'd probably also be surprised if results were very different -- if, say, my afternoon revealed 10% support for the 2 major party candidates with 80% pulling for the Klingon, I'd conclude that I'd missed something on the news (or, alternatively, that I needed a new dentist).


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: