Why not subscribe?

Wednesday, May 04, 2011

Hypothesis testing

http://imgs.xkcd.com/comics/null_hypothesis.png

I just finished teaching statistics again.  The standard curriculum includes a couple of weeks on hypothesis testing, followed by ANOVA and 2 weeks on regression, which also involve testing.

I’m not sure about this. I’m tempted to play this down next time in favor of confidence intervals (error bands around numbers) and simple re-expressions such as the use of per capita income instead of income, and indexing.

Hypothesis testing is important if you are writing a scientific paper for publication, but otherwise not so important. It’s one of those things that they won’t run into THAT much in business – as opposed to confidence intervals, which have a more straightforward relationship to uncertainty.

I’ll get a chance to think this over before I teach stat again.