Feature request -shift the effect size plot for Cliff's Delta to the right side of the observe data

caroline kibat

unread,

May 7, 2020, 9:56:50 AM5/7/20

to estimationstats

Hi Joses,

Is it possible to shift the effect size plot for Cliff's Delta to the right side of the observe data ? Similar to the one we see if we use Cohen's D?

Thanks and take care!

Caroline

Cliff's Delta.png

Cohens D.png

Joses Ho

unread,

May 11, 2020, 5:02:53 AM5/11/20

to estimationstats

Hi Caroline!

Cliff's delta is a measure of ordinal dominance. It's a measure of how likely a random sample from Test will be greater than a random sample from Control. Hence, unlike Cohen's d, there is no formula to convert Cliff's delta back to the original data units. So a Gardner-Altman layout (as you have shown for Cohen's d) is not possible for Cliff's delta.

Hope this helps!

Joses

Lewis Martin

unread,

May 12, 2020, 8:02:04 PM5/12/20

to estimationstats

It might make sense to make the Cliff's delta plot more economical by removing the 'Control' group, since all the information is encoded in the second group anyway, being 'Test minus control'. Since the Cliff's delta panel would now have only one column, then it can be moved nextdoor to the scatter plot, but unlike the mean difference, in this case they would not share a common axis. For anyone interested in doing that you can download the SVG file and it can be done in a minute or so using Inkscape or similar vector graphics programs - I've attached an example

best

lew

--
You received this message because you are subscribed to the Google Groups "estimationstats" group.
To unsubscribe from this group and stop receiving emails from it, send an email to estimationsta...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/estimationstats/6e32205e-e82b-4010-8c4e-cebd44e853bf%40googlegroups.com.