Squared instead of normal pairwise euclidean distance for e-distance calculation #30
Description
Hi!
I very much like the energy distance you guys use in the scPerturb paper and here in the scPerturb package. I have a question about its exact definition, as there seem to be some inconsistencies between different sources.
My understanding is that you guys use the E-statistic as defined in this paper, which is the squared energy distance.
Specifically, according to this paper the mean pairwise distance between sets of points is used to calculate the E-statistic:
E = 2A - B - C, where A, B and C are average pairwise euclidean distances.
However, in the scPerturb paper you say:
"We used the squared Euclidean distance when calculating cell-wise distances.", i.e. you used mean squared euclidean distances for A, B and C.
In your code, too, your default is set to using the squared distance (see here).
Is there any reason you decided to use the squared distance rather than the actual Euclidean distance?
Looking forward to your response!
Also tagging @moinfar here, who looked into this with me