Regression

Regression is a statistical method that models the relationship of a target metric with one or more other metrics, both to quantify the relationship and to enable prediction.

Regression is a step beyond correlation: correlation only says two things move together, while regression tries to build a quantitative relationship saying how much the target metric changes if the input metric changes by so much. Its power is the multivariate case: it can weigh several factors at once and say, holding the rest fixed, what each factor's share is. But regression carries the same correlation trap with more confidence; a well-fitting model still does not prove causation and can make a spurious relationship look definitive. It has two technical dangers too: overfitting (a model that memorizes the noise of past data and breaks on fresh data) and extrapolation (predicting outside the range the model was trained on). Regression is a powerful tool provided its limits are not forgotten.