dama wrote:Varga and Delaney (2000)* has proposed. δ-index for stochastic difference. You wouldn't happen to know if there is a shortcut to calculate that?

Hi, @dama.

The (-1, 0, 1) encoding for your variables makes them available as a dominance matrix, the

mean of which corresponds to the effect size (Cliff's Delta).

There is a linear relationship between Cliff delta and VDA delta (Varga and Delaney, 2000) if we consider that Cliff_d ranges from -1 to 1, while VDA_d ranges from 0 to 1.

How do I get VDA_d from Cliffs_d:

VDA_d = (Cliffs_d + 1) / 2I don't think it is important which of the two to report, I consider that it is possible to calculate one from the other.

If you ask the question, which is better, I personally prefer Cliff's delta, since it scales between -1 and 1, and if you take it as an absolute value, between 0 and 1, with 0 = no effect (overlapping data) and 1 effect stronger (no overlap between data).

It can be said that the VDA delta is calculated so that no effect produces a value of 0.50.

A value of 1 indicates the complete stochastic superiority of one group and a value of 0 indicates the complete stochastic superiority of the other group.

I hope I've answered your question, but if you want to look at some R functions to use with the Rj module you can take a look at Marco Torchiano's "effsize" package:

https://github.com/mtorchiano/effsize/tree/master/R

Cheers,

Maurizio