Do f58r/f58v show Currier languages evolving?

[ReneZ]

This shows the fraction of "ed" bigrams over the pages. The light grey crosses are the zodiac pages, which we know are in the right order, and should realistically be of similar subject matter. Here we see the "ed" bigram count increase from zero to something distinctly non-zero.
The zodiac pages consist of circular text and labels, and the vast majority of these bigrams occur in the circular text, where also the increase is being observed. However, the increase is also seen in the labels. In the early signs, they do not occur. In the later signs they start to appear.

I guess the big question is whether there is a different feature that decreases as the ed bigram count increases.

Indeed, this is not yet clear. It is not even clear to me that "ed" is a unit in any form, or the ending of one unit followed by the start of another unit.
I attach the fractions plot (per page) for other bigrams including "eo", this time in 'standard Cuva'. It confirms Marco's observation about "eo" in f58r/v vs. the pharma pages.

   

Cuva HO is Eva qo, SO/SE are cho/che and ZO/ZE are sho/she.

I also attach another version of my PCA plot, where I made the following changes:
- It uses Cuva instead of Cuva1
- Each cross is a folio, not a page
- It uses the standard PCA algorithm, not my old 'hack'.

This should be very similar to Marco's plot. Not sure how 'close' this should be considered to be.

   

The blue cross near the green ones (Astro/Cosmo) is f58.