Every simulator from the course in one place, and eight calculators that show their working rather than just their answer — so you can check a computation and still learn something from it.
Each one appears in its own volume too. Ten minutes here is worth an hour of reading — particularly the confidence-interval simulator, which is the fastest cure for the standard misinterpretation.
Draw samples from a skewed population and watch the distribution of their means go normal anyway. Change n and watch the spread shrink by √n.
Run a hundred studies from a known population. About five intervals miss. The probability belongs to the procedure, not to any one interval.
Move the true effect and the sample size and watch the four quantities trade against one another. Set d = 0.3 with n = 30 and see how little power the typical study has.
Drag your test statistic and watch the tail area. Change df and watch t approach the normal curve.
Set r to .30 — a typical published effect — and look at the cloud. Then hide the value and guess it a dozen times to calibrate your eye.
Separate the group means and watch F rise; widen the within-group spread and watch it collapse. F is a ratio of two kinds of variability, and nothing more.