11 of 14 1 Points Which of the following are reasons employ the data-splitting technique. (Pick all of the answers that are correct.) A. Guard against overfitting by testing the accuracy of estimated Prediction Intervals. B. When differencing does not work. C. When your original model does not produce Normally distributed residuals. D. Help to choose between two or more valid models. Question 12 of 14 1 Points When deciding which data to exclude in order to create the test set, the rule of thumb is: A. Approximately 25%-50% of the oldest data. (In other words, the first appx 25% of the data.) B. Randomly select approximately 25%-33% of the data. C. Approximately 25% of the most recent data. (In other words, the last appx 25% of the data.) D. The middle third of the data. Reset Selection Question 13 of 14 1 Points When evaluating the accuracy of the Prediction Intervals, you need to: A. Compare the Prediction Intervals from the model built on the full dataset with the predictions from the model built from the Training Set. B. Compare the Prediction Intervals from the model built off the Training Set with the actual observations for those time-periods