confounding variable
In plain wordsA hidden third factor that influences two things at once and makes them look linked. Ice-cream sales and drownings both rise in hot weather; neither causes the other.
A common cause of a treatment (or exposure) and an outcome which, left unaccounted for, distorts their apparent relationship — manufacturing an association where the treatment has no effect, or hiding or exaggerating one that exists. Observational data are full of them: sicker patients receive more treatment and also fare worse; power users both adopt new features and stay longer. Randomization removes confounding by breaking any link between confounders and treatment assignment. In observational studies one must measure and adjust for a sufficient set of confounders — identified from a causal graph by the back-door criterion of the do-calculusdo-calculus, n.: A set of mathematical rules, devised by Judea Pearl, for working out the effect of deliberately changing something using only observational data plus a diagram of what causes what. — while avoiding adjustment for mediators or colliders, which can create bias. Unmeasured confounding cannot be ruled out from data alone. In machine learning, confounders produce spurious correlationsspurious correlation, n.: A link in data that fools you: two things move together, yet neither drives the other, because a hidden shared cause, the way the data were collected or luck produced the pattern. AI models readily latch onto such links. that models exploit and bias offline evaluation of recommender systemsrecommender system, n.: Software that guesses which songs, products, posts or jobs you are most likely to want and shows those first, as when a video app lines up the next clip or a shop suggests an add-on. and policies from logged data.
Prompt length seemed to hurt accuracy, but task difficulty was a confounding variable driving both.
Origin
From Latin confundere, 'to pour together, mix up'. Ronald Fisher's The Design of Experiments (1935) used 'confounding' for effects that a design cannot separate, and epidemiology developed the modern causal sense in the mid-20th century; Judea Pearl's back-door criterion (1993) gave a graphical test for adequate adjustment.
Related
- causal inferencecausal inference, n.: Working out whether one thing actually causes another, rather than merely occurring alongside it — for example, whether a new feature made users stay longer or just happened at the same time.
- spurious correlationspurious correlation, n.: A link in data that fools you: two things move together, yet neither drives the other, because a hidden shared cause, the way the data were collected or luck produced the pattern. AI models readily latch onto such links.
- Simpson's paradoxSimpson's paradox, n.: A statistics trap in which combining groups reverses what each group shows: a treatment can do better than another for men and for women separately, yet worse when everyone is counted together.
- do-calculusdo-calculus, n.: A set of mathematical rules, devised by Judea Pearl, for working out the effect of deliberately changing something using only observational data plus a diagram of what causes what.
- structural causal modelstructural causal model, n.: A precise description of what causes what in a system — a diagram of arrows plus equations — that predicts what would happen if you intervened, not just what tends to occur together.
- correlationcorrelation, n.: A measure of how closely two quantities move together, from −1 (one rises as the other falls) through 0 (no straight-line link) to +1 (they rise together). It does not show that one causes the other.
Mentioned in
correlationcorrelation, n.: A measure of how closely two quantities move together, from −1 (one rises as the other falls) through 0 (no straight-line link) to +1 (they rise together). It does not show that one causes the other., causal inferencecausal inference, n.: Working out whether one thing actually causes another, rather than merely occurring alongside it — for example, whether a new feature made users stay longer or just happened at the same time., spurious correlationspurious correlation, n.: A link in data that fools you: two things move together, yet neither drives the other, because a hidden shared cause, the way the data were collected or luck produced the pattern. AI models readily latch onto such links., do-calculusdo-calculus, n.: A set of mathematical rules, devised by Judea Pearl, for working out the effect of deliberately changing something using only observational data plus a diagram of what causes what., Simpson's paradoxSimpson's paradox, n.: A statistics trap in which combining groups reverses what each group shows: a treatment can do better than another for men and for women separately, yet worse when everyone is counted together., structural causal modelstructural causal model, n.: A precise description of what causes what in a system — a diagram of arrows plus equations — that predicts what would happen if you intervened, not just what tends to occur together.