#mape

Question 1: Single choice

A model predicts the monthly revenue of business units of very different sizes. On the test set it reaches $R^2 = 0.92$ but a mean absolute percentage error (MAPE) of 40%: it is accurate for the large units and poor for the small ones. What best explains how both numbers can hold?

One of the metrics must be miscomputed, because a high $R^2$ implies a low MAPE1
$R^2$ is dominated by squared errors on large units; MAPE weighs every unit's relative error equally2
A MAPE of 40% means the model explains only 60% of the variance, so $R^2$ must be wrong3
$R^2$ ignores the large units because it is computed on percentages4

Timed test