The proof and measurement of association between two things

Abstract

Attempted to study the scientific correlation between two things. Any correlational experiment can only be regarded as a sample and presents a certain amount of accidental deviation from the real general tendency. Accidental deviation can be measured by the 'probable error.' Accidental deviation depends on the number or cases, and on the largeness of existing correspondence. Probable error varies according to the method of calculation. While the number of Ss helps to reduce accidental deviation, it has no effect upon systematic deviation, except that it indirectly leads to an augmentation. Therefore, the number of cases should be determined by the principle that the measurements to be aggregated together should have their error brought to the same general order of magnitude. Suggests that probable errors must be kept down to limits small enough for the particular object of investigation to be proved.