desperately needed: statistical literacy

What do these confidence intervals show?This piece on Vioxx from NYT is interesting for a few reasons – none more so than for how crucial the interpretation of statistical evidence can be. Vioxx trials will involve people offering both the naïve intuition about confidence intervals (if they overlap, there's no difference; good for Vioxx) and the correct one (you need the full variance-covariance matrix to perform a hypothesis test).

I’ve long argued that basic statistics should be required in high school instead of either geometry or algebra II. It wouldn't make much difference in the case of tricky interpretations like this one, but I do think that people would be better served by a working knowledge of basic statistical theory more than by sines or Cramer’s rule.

No comments: