"The p-value reported by R is also very, very low, which seems good, but remember all this really means is “you’d have to be crazy to think a flat line fit better than one with a slope."
— Cosma Shalizi, Testing Parametric Regression Models with Nonparametric Smoothers
