A database contains data on diamond prices and quality. Parametric methods deal with colour only as a category. But since diamond colours are ordered with D as the best and later letters labelling lower-quality colours, it makes sense to ask whether prices change consistently with colour quality. We can rank the colours, assigning rank 1 to colour D, 2 to colour E, and so on. Of course, these ranks have no measurement units, so a least squares regression would make no sense. But it is appropriate to use Kendall’s tau to assess the association between Price and Colour Rank.
QUESTION: 

a. What is Kendall’s tau for these data, and what does it mean?

b. That doesn’t make sense. How can we explain this apparent anomaly?