The phrase Kendall Tau is not typically introduced in your basic statistics class. But if you were performing a correlation for your  dissertation defense- make sure you don’t need one. As a matter a fact, do me a favor…If you are working with a correlation right now…Classify the variables in your correlation as one of the following : interval, ratio, nominal, or ordinal. If your variables are ordinal a Kendall Tau is more appropriate than a Pearson.   The Kendall Tau is a non-parametric correlation for ordinal variables. Ranked Correlations for ranked data!

The Pearson’s correlation coefficient is more appropriate for continuous variables (interval or ratio). When you are studying linear correlation, your variables should be aligned with a normal distribution or at the very least the linear family. And this is my perspective, but count data is not part of the linear family.

There are other non-parametric correlation tests. Know your variable type.

-Moore to Follow Amy

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.