Calculates Kendall's K statistic for a 1D binned dataset between the axis and data components, or between 2 primitive 1 dimensional objects.
The K statistic is a distribution-free test for correlation - see Ponman, MNRAS, 201, p769, 1982.
In the absence of any correlation the K statistic is approximately unit normally distributed (so long as there are at least 10 data values), so a value K = 3.0 constitutes three sigma evidence for positive correlation. Negative values imply a negative correlation.
Bad quality data are excluded from the statistic. Variance is not used.