All experimental analysis should rely on multiple datasets, but seldom are the consistencies between datasets carefully quantified. Morillas et al. examine this question for color difference perception where there are naturally large levels of inter-observer and inter-experiment variation. In their work, consistency is defined as similarity of the magnitudes of color difference judgments across different research studies. They develop and apply a “fuzzy analysis” technique that defines statistically when data sets have color differences that are “similar” while having locations in color space that are “not far apart”. The fuzzy analysis is necessary in this application since precise metrics fail due to ill-defined tolerances (the point of much of this research). Ultimately the technique allows for a type of smoothing of the various data sets by collecting the similar data which, in turn, allows the creation of more stable and meaningful mathematical color difference equations. This analysis should have many useful applications in the future and perhaps an extension to bi-directional experiments in which changes of perception, or other measurements are measured for two different directions of stimulus, or adaptational, change.
You must log in
to add comments.