In statistics, the degree to which the probability of drawing a wrong conclusion from the test result is not seriously affected by moderate departures from the assumptions implicit in the model on which the test is based. [L. robustus, hale, strong, fr. robur, oak, strength]
