Attenuation: How Noisy Measurement Shrinks a Correlation
The deepest node of the sub-domain and its payload for the rest of the path. Random measurement error in either variable drags an observed correlation toward zero, so a small reported association can be a real association measured badly. It is the honest reading of the seed paper's weakest numbers — 41 of its 72 measures came in under 5% heritability, and its formant heritabilities sit at 8–9% — alongside its own admission of formant estimation error.
Questions this Concept answers
- Why does a weak observed association have two explanations that look identical from the outside?
Correction for Attenuation in Classical Test Theory
The correction for attenuation, derived from classical test theory (Spearman, 1904), estimates the correlation between the true-score (error-free) variables underlying two measured variables by divid…