When it is clearly seen
That the sample is extreme,
Then p is small,
Reject the null.
There's a hippopotamus.
He's got a hypothesis.
His null specifies a parameter,
That he's still skinny, no change in his diameter.
And if there is a change, that he's gotten fat from all those fries,
Well that's his alternative, a deviation from his hypothesized size.
Hoping for the best, he checks his p-value,
The probability of having extreme data, if the null is true.
Because his p-value is small, he comes to realize,
He needs to reject his null, and start to exercise.
But when he weighs himself, he soon comes to see,
That there was a type 1 error and he really is skinny.
His null was really true, but he had rejected it,
Because his p-value was below threshold and therefore statistically significant.
But was the change practically significant? He didn't think so.
For a tonne-weighing hippopotamus, a 10 pound change is too low.
So although weighing a tonne may not be skinny for me or you,
The hippopotamus thought so and shouted "Yoohoo!"
Chance explanation?
The value is too extreme
That's significant
Dear Karl, thanks for correlation
Even though it may not be causation
Your one greatest hit
Chi-square, that was it
From "normal" we have liberation
There was a young Student at Guinness
Who studied a beer as his business
His small sample sizes
Were full of surprises
With Fisher, the "t" he did finish
