[split] EVA-y as Latin "-us" etc.

[ReneZ]

Every statistical computation is based on a number of assumptions. These assumptions are not always (or even rarely) spelled out. The result of the computation, and the conclusion, usually say much about the correctness of the assumptions.

When doing entropy calculations, there is a complicated set of assumptions about what constitutes the Voynich alphabet and which are the distinguishing character shapes. There is also an assumption that spaces are word separators.
The result is then compared with known plain texts, but:
- these plain texts are usually fully edited
- with any abbreviations expanded
- the Voynich MS is handwritten and may have any number of undetected errors
A final assumption is therefore that these points do not greatly affect the outcome

The error from the main assumption (of the alphabet) can be mitigated by redoing the analysis with different alphabets. Given that the resulting differences are very significant, one may have faith in the validity of the anomalously low conditional character pair entropy.

With the question whether word-final y represents word-final -us the situation is far simpler and there are basically only two assumptions:
- the apparent spaces in the MS text are word separators
- there is one symbol that looks like y that can be distinguished from all other symbols
Notably, the rest of the Voynich alphabet does not matter.

Both assumptions are actually part of the proposed equivalence of y with -us. If either assumption were invalid, the thesis disappears. Therefore, there is no issue, and the enormous difference between appearance of y in the Voynich MS and in known Latin texts is clear.

Again, this is true because of the length of the text. Another extreme is the Dorabella cipher, which is so short that statistics cannot be relied upon in any useful manner.