Where does correlation collide?

In statistics, correlation is the strength and direction of the linear association between two quantitative variables, summarised as a single coefficient that runs from -1 through 0 to +1: it says how tightly the paired values track a straight line, and nothing about why. Pearson's 1896 Mathematical contributions to the theory of evolution, III: Regression, heredity and panmixia gave the product-moment coefficient its modern form, building on Galton's earlier study of regression toward the mean. The referent is a symmetric, unitless summary of co-variation in a sample, estimated from data, and it is not a dependence between quantum measurements, a marker that predicts immune protection, or the philosophical relation of cause to effect.[1]

Unit

A unitless coefficient between -1 and +1, estimated from paired data

Recovery means

Not applicable

Measured by

Computed as the covariance of two variables divided by the product of their standard deviations (Pearson's r), or from ranks (Spearman's rho)

In quantum physics, a correlation is a statistical dependence between the outcomes of measurements made on separate parts of a system: when two particles are entangled, the result recorded at one detector is related to the result at the other in a way that can exceed anything a shared classical cause could produce. Bell's 1964 On the Einstein Podolsky Rosen paradox derived an inequality that any local hidden-variable account of such correlations must satisfy, and which quantum mechanics is predicted to violate. The referent is a joint distribution over measurement outcomes constrained by physical law, not a coefficient fitted to a sample, a vaccine's correlate of protection, or a metaphysical link between cause and effect.[2]

Unit

A joint probability distribution over paired measurement outcomes

Recovery means

Not applicable

Measured by

Estimated from coincidence counts across many entangled pairs and tested against a Bell inequality, such as the CHSH bound

In vaccine immunology, a correlate of protection is a measurable immune marker, most often an antibody level, that reliably predicts whether a vaccinated person is protected against a disease: it is the quantity a regulator can check in place of running a full efficacy trial. Plotkin's 2010 Correlates of protection induced by vaccination distinguished a mechanistic correlate, which does the protecting, from a non-mechanistic one, which only tracks it. The referent is a biomarker validated against clinical outcomes and expressed as a protective threshold, and it has nothing to do with a statistical coefficient, a quantum dependence, or the philosophy of causation.[3]

Unit

A protective threshold of an immune marker, such as an antibody titre

Recovery means

Not applicable

Measured by

Established by relating a measured immune response to observed protection, often as the titre above which most vaccinees are protected

In the philosophy of science, correlation names an observed statistical association between events that is carefully held apart from causation: two things may vary together because one causes the other, because a third factor causes both, or by chance, and the association alone cannot decide which. Reichenbach's 1956 The Direction of Time made this precise with the principle of the common cause, that a correlation between two events with no direct causal link points to a common cause that screens them off. The referent is an epistemic relation used to reason about cause, symmetric and non-committal on its own, not a fitted coefficient, a quantum dependence, or a validated immune marker.[4]

Unit

An observed association treated as evidence about, but not proof of, causation

Recovery means

Not applicable

Measured by

Assessed by whether an association survives controls for confounding and supports intervention, not by the strength of the association alone