To coin a paradoxical phrase: least-squares linear regression is the best of statistics, it is the worst of statistics. It is the hest of statistics because it is often appropriate in practical situations and is theoretically tractable; it is the worst of statistics because the assumptions upon which it is based (1-4) may be inapplicable to the problem at hand and because it is fatally prone to effects of "wild" outlier data.
展开▼