Correlation coefficients and correlation ratios as estimates of observer weights in multiple‐observation tasks

Advantages of correlational methods are described for estimating the relative reliance or weight observers give to different elements of a complex display when performing auditory or visual discriminations. Correlational methods differ from other such methods by providing a least‐squares estimate of weights rather than a maximum‐likelihood estimate based on a presumed model. The main benefit is that these methods can be generally applied in cases where the statistical properties of elements are unknown, where correlations exist between elements, or where it is necessary to evaluate decision models that are inherently nonlinear.