Treatments never look at the whole population. That is why parametric statistics exist. It is impossible to look at every combination of 16Mx16M colors. We compute sample sizes. My guess is that the number would be quite small. My author's test is what one would use prior to a final accessibility audit. To be absolutely certain one could proceed as follows: Determine S = the set of colors used on the page (excluding images) Select a random colors from the sets I described that are not in S. Then do the test. Two colors is all we need because there are no accidental conflicts that could give false positives. Wayne
Received on Sunday, 2 April 2017 20:05:55 UTC