In the motivating statistical experiment, it may make sense to choose the assignments in a more balanced fashion than assigning individuals independently. For example, if there are 10 subjects, you might choose randomly from among the ways to assign 5 to each treatment rather than risking the assignment of 9/10 to treatment A by luck.

The formula you give for the standard deviation differs from the one in Schilling’s papers by a factor of pi/sqrt(6).

The distribution about the mean is not normal. The tails drop off only roughly exponentially (for a while).

]]>if there were a collection of 100 surprising math facts about the real world every scientists and engineer should know, this would surely be on the list. ]]>

`1/m.log10(1/0.5) 3.321928094887362`

, but `math.log`

means ln. So what’s the base of your log? 10, e, 2? ]]>