HI @devonorourke,
Thanks for tagging me in this discussion! This is an area I'm getting into and getting excited about; the referenced post was awesome.
I think you're looking for two separate functionalities: statistical test and visualization. I think they're closely linked, but slightly different things.
As far as I understand, permadisp is the statistical test that measures whether the dispersion of a group of samples in ordination space is different between groups (essentially a measure of variance). I usually use the scikit-bio implementation, but the result is essentially the same. The permadisp function is statisitcal test that gives a psuedo-F and permutative p-value for whether or not there's dispersion between the two groups or more groups.
Visualization is a second issue and is maybe more subjective, but that's the plot. QIIME (again AFAIK, Im not a developer and do most of my own analysis in python because I like control) doesn't have a way to visually show the centroids for clusters. Personally, I think I'd rather see the centroid as a cloud with a variation limit rather than the center point (like transparent, or something).
I think the steps to getting there for an initial prototype might be the following:
- Calculate a median position along each PC using something like beta rarefaction
- Somehow merge the median PCoA with the main PCoA using something like bi-plots
- Coerce the shiny new biplots into emperor
- ????
- Profit
My list of steps is probably a vast oversimplification and I'm missing something obvious, but I think that would be my first pass to try it. Probably in the underlying python architecture. Although @yoshiki can probably give limitations of emperor better than I can.
I don't have the bandwidth right now to play around this. Maybe over the summer when everyone here goes off on vacation 
, Ive got a confusing number of hours of daylight, and theoretically more free time. (Although that feels like a dangerous promise.)
Best,
Justine