This is why statisticians do a ‘correction for multiple comparisons’, a correction for ‘rolling the dice’ lots of times. This, like correcting for clustering, is particularly brutal on the data, and often reduces the significance of findings dramatically.
