The best linear prediction rule (i.e., model) is the one that has the:
- Greatest squared error when predicting using that rule.
- Greatest cumulative error when predicting using that rule.
- Greatest total error when predicting from the mean.
- Least squared error when predicting using that rule