Ah! This extreme case is interesting. Here is a situation where the MAE is basically zero (VERY little 'error of the model') but the r value indicates 'no relationship.'
So, measuring the 'error of the model' is not the same as measuring the 'strength of the relationship'
