Character entropy of Voynichese

[Anton]

I computed the single character entropy of all word-initial characters by simply counting the word-initial characters, and computing their frequency distribution.

But you compute frequencies across all the text, or across subsets of first/last letters only?

In the following example:

Code:

this is a test message

the frequency of 't', for the word-initial, would be 2 or 3?

EDIT: Ah, OK, surely the former. Otherwise it would not make sense. I wonder what would be the respective result for second letters.