Sunday, March 06, 2011

Statistics: How not to use them

Two recent articles on how statistics can be (mis)used to mislead people about the thing being discussed. This really gets at the heart of one of my pet peeves about news: just because you can toss a number, ratio, or percentage out does not make it helpful to understanding the issue in question.*

Here are two cases in which statistics may actively obscures the issue:

1) A general post on When is an Average Not an Average? It does a great job pointing out why averages don't actually mean anything in a large number of situations.

2) A much more controversial post, wherein the author looks deeply at a statistic that failed the sniff test, and discusses the statistical implications of the fact that concentration of the incidence within a small group distorts the statistic being applied to the entire population.

Interesting to the math geek in me.

*On a related note, a group of grad students I took a Finite Element Theory class with had a running joke that the analyses the presenters showed us had to be right because they had pretty color pictures of the results. The implication of course being that no matter how pretty the analysis/picture/statistic looks, it's only as good as the foundational assumptions on which it is built. Having a number/picture/whatever does not make the analysis correct. Different point, but related.


Post a Comment

<< Home