When the data is not normally distributed, which fit metric is preferable to use and why? MAE because it uses the absolute value of the residuals, so it's more accurate MAE because it squares the differences between the actual and predicted values, amplifying the effect of larger residuals RMSE because it uses the absolute value of the residuals, so it's more accurate RMSE because it squares the differences between the actual and predicted values, amplifying the effect of larger residuals