The Art of Statistics: Learning from Data

by David Spiegelhalter

A P-value is a measure of the incompatibility between the observed data and a null hypothesis: formally it is the probability of observing such an extreme result, were the null hypothesis true.