Meaning of 'Regression'

Meaning of ‘Regression’ in Linear Regression


What is the meaning of “Regression” in Linear Regression ?

Regression = Going Back

Regression has a very specific historical origin that is different from its current statistical meaning.

Sir Francis Galton (19th century), cousin of Charles Darwin, coined 🪙 this term.

Observation:
Galton observed that -
the children 👶 of unusually tall ⬆️ parents 🧑‍🧑‍🧒‍🧒, tended to be shorter ⬇️ than their parents 🧑‍🧑‍🧒‍🧒,
and children 👶 of unusually short ⬇️ parents 🧑‍🧑‍🧒‍🧒, tended to be taller ⬆️ than their parents 🧑‍🧑‍🧒‍🧒.

Galton named this biological tendency - ‘regression towards mediocrity/mean’.

Galton used method of least squares to model this relationship, by fitting a line to the data 📊.

Regression = Fitting a Line
Over time ⏳, the name ‘regression’ got permanently attached to the method of fitting line to the data 📊.

Today in statistics and machine learning, ‘regression’ universally refers to the method of finding the
‘line of best fit’ for a set of data points, NOT the concept of ‘regressing towards the mean’.

images/machine_learning/supervised/linear_regression/line_of_best_fit.png



End of Section