Essentially, all models are wrong, but some are useful.
p. 424, - Empirical Model-Building and Response Surfaces (1987)