News for hackers. Stuff that matters.
Re:Skeptics discount science by casting doubts on scientist ... by yummyfajitas on Wednesday September 15, 2010 @09:56PM (#1696577)
While the graph is nice, error bounds would be extremely useful. Comparing two groups based on a sample size of 5 (per group) is an extremely tricky thing to do.

It's even trickier when you try to rule out systematic errors by reducing your sample size further. For instance, the blog mentions one systematic error caused by D->R transitions. He then claims this is ruled out by analyzing two R->R and one D->D transition, ignoring the fact that 2 of those three transitions were based on fluke circumstances themselves (specifically Kennedy->Johnson and Nixon->Ford).

There are no comments on this comment yet.