This applet allows you to specify two groups of a quantitative response (in either stacked or split format) and then produce the randomization distribution.

Setting Number of repetitions to 1 shows the groups being shuffled. Setting Number of repetitions between 1 and 30 shows the groups as they are formed.  Number of repetitions greater than 30 turns off the animation.

The hypothesized group 1 effect is the "benefit" you assume from being in group one. Nonzero values are subtracted from group one before the data are re-randomized and then added back on to the observations assigned to group one.

You can always view the original data by opening and closing the data entry windows.

Scrolling down in the page seems to work better with Firefox than with IE.  Sometimes you also have to click in the whitespace surrounding the main window to be able to scroll.