Regression & Correlation
Galton (1886) and Pearson (1890s) found ways to measure how strongly two quantities move together; regression and correlation became the basic tools for finding patterns in data.
Open in the interactive tree →Francis Galton, studying the heights of parents and children, reported in 1886 that extreme traits are only partly inherited and children drift back towards the average, which he called regression, and in 1888 he defined an index of correlation. Karl Pearson put the mathematics on firm ground in papers from 1893, including the correlation coefficient and the chi-squared test of 1900, a way to judge whether a pattern could be chance; both men also promoted eugenics. R. A. Fisher refined regression in 1922-25 and joined it to his significance tests, and regression is still the basic way to estimate how one thing depends on another, from economics to medicine.
Prerequisites
- Statistics & Least Squares1809Regression extends Gauss's least squares to pairs of measurements.
- Evolution by Natural Selection1859Galton studied how traits are inherited in the wake of Darwin.
Unlocks
- Field Trials and Randomization1925Fisher's analysis of experiments extends regression and correlation.
- Significance Testing (P-Values)1925-1933Pearson's chi-squared test and correlation preceded Fisher's tests.
- Modern Evolutionary Synthesis1942Fisher's population genetics used Galton and Pearson's correlation between relatives.