I read an interesting article on variability The Most Dangerous Equation that was published in American Scientist. It discusses the fact that variability in measurement decreases with the square root of the sample size (which is pretty slow), and then gives examples of real-world implications of this fact. For example, Prof. Wainer discusses how small schools maybe come out "better" in most studies because their variability is higher: not only are they among the best schools, they are also among the worst.

No comments: