0
$\begingroup$

I have two samples A and B in excel and I want to use the vargha and delaney effect size in R. VD.A (A, B) gives me the result as. 0.170 (large)

My question is how it can be large with this value because the large value in vargha effect size begins with 0.7 and above.

Improve this question
$\endgroup$

1 Answer 1

Reset to default
0
$\begingroup$

If you had used VD.A(B, A), the result would be 0.83. With VDA, no effect is 0.50 and 0 or 1 is complete stochastic dominance of one group over the other. This can be a little disconcerting at first, but it makes a lot of sense for this statistic. VDA is the probability that an observation in one group will be larger than an observation in the other group. So the result of VD.A(A, B) is equal to 1 - VD.A(B, A).

It may be helpful to compare this to Cliff's delta, which is a linear transformation of VDA, where 0 is no effect, and -1 or 1 is complete stochastic dominance.

Improve this answer
$\endgroup$
6
  • $\begingroup$ thanks a lot. Can we use Cliff's d for data that is not normally distributed? $\endgroup$
    – Khan
    Commented Oct 21, 2019 at 12:38
  • $\begingroup$ You can use Cliff's delta any place you can use VDA. (Which I think is just needing two groups of observations of ordinal or numeric data). Cliff's delta is literally equivalent to VDA just translated from a scale of 0 to 1 to a scale of -1 to 1. $\endgroup$
    – Sal Mangiafico
    Commented Oct 21, 2019 at 12:45
  • $\begingroup$ thanks a lot once again $\endgroup$
    – Khan
    Commented Oct 21, 2019 at 12:56
  • $\begingroup$ Hi Sal, the Cliff's delta test gives the results as, what does it mean? -0.8337751 (large) $\endgroup$
    – Khan
    Commented Oct 21, 2019 at 13:01
  • $\begingroup$ The negative sign means that the second group has higher values than the first group. This is the common convention for Cohen's d and other comparisons of two groups. Just be careful that you know what order R is using for the groups. For example with levels(Group_var).... Both CD and VDA assess the same thing: the probability of an observation from one group being larger than an observation in the other group. VDA just reports this as an actual probability (0.50 is no effect), and CD translates this to a -1 to 1 scale (0 is no effect). $\endgroup$
    – Sal Mangiafico
    Commented Oct 21, 2019 at 13:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.