I’m putting the final touches to the second article on SWM to be submitted soon to a statistical journal.

In it I expand on the mathematics behind SWM to show why one should never try to compute arithmetic averages from ratios of the a/b form and similar forms. There are many reasons against doing this: Cauchy Distributions, additivity arguments, Hölder’s Inequality concepts, etc.

The paper kills for good the notion of arithmetically averaging correlation coefficients and other type of ratios frequently encountered in statistical analysis.

Part of this will be part of my presentation on Data Mining for Clinical Trials before Fundación de Investigación.

About these ads