You have a set of x, y pairs of numeric data (ratio values) that you want to assess for correlation. they appears to be a relationship -- overall, y scores tend to go down as x go up, although not linearly -- but the data are noisy so this is not true for all x, y pairs. what is your best option?