Evidence for 2 rules linking words in Herbal A->B

[npcompl33t]

Your lists of words are almost all gallow words, and the strings that you consider to be of significance,   chy, edy, chdy   are suffices to the gallow words ( i.e. the segments that comes after the gallow character ).

Virtually every voynich word can be decomposed into a suffix that follows a gallows character, and to allow suffixes to freely move like this permits you to change almost any voynich word into another word. 

As much as possible, the suffix of the word needs to remain constant, because considering suffixes swappable will very quickly boil most voynich words down to just a handful.

The type of frequency analysis you are describing is indeed something I am doing, but it alone is not sufficient. For instance, aiin can basically be swapped out with any suffix and the  frequencies match. This does not mean that qokedy and qokaiin are the same word. It means qok occurs at roughly the same frequency across the manuscript.

To establish a rule, words that appear mostly if not entirely in one section need to be mapped into words that are also rare or nonexistant, and all words that could be created by a rule need to be check to see if the accounting all lines up -- would that rule generate far more of a particular word than we actually see? 

Several of the words you mentioned i would also not consider suffixes, even though they can appear in another word. It would be like calling 'our' a suffix in English because pour, four, tour, flour all exist.
'ire' would be another englsh word: tire, fire, sire, quire, etc. 

If you were to guess that ire and our are suffixes, four and fire are the same word, because f + ire or f +our.

So word endings does not automatically = suffix, even if many words seem to share it. They are not freely changeable with different forms, like how 'hike' can take both -ing, -ed, or -er to be hiking, hiker, hiked.

Part of the challenger here is differentiating between the two, and i think frequency analysis is definitley a part of it, but shared frequency alone is not sufficient, in particular because quickly you will find that to make everything work out some endings, like -edy, need to function as a type of abbreviation, with multiple different words mapping into them. So you have to do the accounting of transforming all the words using all the rules, adding up all the generated words, and making sure everything still lines up even if mutiple rules would turn different words into a single word.