Monday, March 22, 2004


I spotted an interesting discussion of statistics on a business blog called Synergy Fest. The site pointed out inconsistencies in logical thinking, especially with probabilities. The entry focuses on Bayesian statistics and raises two cases you are likely to get wrong.

The cases did not impress me so much as the reminder that we should not flaunt numbers without understanding assumptions and facts behind them. That is the problem with factoids. They sound good, but frequently they mislead or are wrong. Few challenge them, however, and they become part of assumptions. No wonder we get off on the wrong track so often.

It is hard work to look at how someone arrived at numbers being used -- hard but necessary work. The quickest way to vet statistics, however, is to look at sample size, whether the sample was random and the percentage of error. That should be standard in PR. We can save the world meaningless junk surveys or stupid interpretations that fall well within a margin of error. Along this line, it is disgraceful how journalists are writing that Kerry is ahead of President Bush by three points or Bush ahead of Kerry when three points mean nothing given the variability of the survey. It's meaningless handicapping.

There is enough bad data in the world to confuse everyone. There is no need to add to it.


Post a Comment

This page is powered by Blogger. Isn't yours?