noise–effect–number triangle

Terms from Statistics for HCI: Making Sense of Quantitative Data

Three factors that interact to influence the statistical power of an experiment or study: the amount of noise, the effect size and the number of subjects or trials. You are likely to get a positive result if the effect size is a lot greater than the size of the noise divided by the number of subjects/trials (effect >> noise/number). Hence a smaller noise, a larger effect or more subjects or trials all lead to increased power. Techniques for increasing power can target one or more of the elements of the noise-effect-number triangle.

Defined on page 107

Used on pages 107, 108, 109, 110