The concept of "regression" as a statistical phenomenon has its origins in the field of genetics, popularized by Sir Francis Galton in the late 19th century. His seminal work, "Regression towards mediocrity in hereditary stature," laid the groundwork for understanding how extreme characteristics in parents are not always fully passed on to their offspring. Galton's observations and subsequent quantification of this trend were pivotal, leading to the invention
of linear regression analysis and influencing much of modern statistical modeling. The term "regression" has since expanded beyond its initial biological context to describe various statistical tendencies.
Galton's Observations on Hereditary Stature
Sir Francis Galton meticulously measured the heights of hundreds of people, observing a consistent pattern: the offspring of parents with extreme heights tended to have heights closer to the population average. He termed this "regression toward mediocrity," noting that characteristics in offspring regress toward a central point, which has since been identified as the mean. Galton quantified this trend, stating that "the average regression of the offspring is a constant fraction of their respective mid-parental deviations." This meant that the difference between a child and its parents for a given characteristic was proportional to the parents' deviation from the typical population. For height, Galton estimated this coefficient to be about 2/3, suggesting that an individual's height would measure around a midpoint that was two-thirds of the parents' deviation from the population average.
Galton's explanation for this phenomenon was rooted in the idea that a child inherits genetic material not only from their immediate parents but also from a broader ancestry. He posited that the further back one traces a genealogy, the more numerous and varied the ancestors become, eventually resembling a random sample from the general population. While modern genetics offers a more detailed understanding of inheritance, Galton's insight highlighted the statistical tendency for traits to revert towards the mean across generations, even if the mean itself is influenced by selection pressures.
The Galton Board and Visualizing Regression
To further illustrate his findings, Galton also published results using a simpler, more tangible example: pellets falling through a Galton board. This device demonstrates how individual random events (the pellets bouncing off pins) collectively form a normal distribution, centered directly under their entrance point. If these pellets were then released into a second gallery, representing a second measurement, Galton posed a reverse question: "From where did these pellets come?" The answer, he found, was not "on average directly above," but rather "on average, more towards the middle." This is because there were more pellets above the middle that could wander left than there were in the left extreme that could wander right, inwards. This visual representation helped to solidify the understanding of how random variation contributes to the regression effect.
Evolution of the Term and Modern Applications
Since Galton's initial work, the term "regression" has been adopted in various statistical contexts. While it originated from observations in genetics, modern statisticians now use it to describe phenomena like sampling bias, which may have little direct connection to Galton's original biological studies. The analytical tools developed from his work, particularly linear regression analysis, have become fundamental in statistical modeling across numerous disciplines. This evolution underscores the enduring impact of Galton's pioneering observations, transforming a specific biological insight into a broad statistical principle applicable to diverse fields, from economics to social sciences, whenever relationships between variables are being estimated.
