Probability distribution of EVA letters within a random VMS word

There is a really nice blog post here:
Letter Distributions in the English Language and Their Relations -- Tim Hargreaves
You are not allowed to view links. Register or Login to view.

I didn't do that blog post justice and full description of this method can be found there, but i ended up with this
it's a nice little visualization of where the EVA-letters are likely to be found in a vord.

If we select a random word from the ZL-3a transcription of the vms,
then for the EVA-lettters in that word, these graphs show where in the word that letter is likely to occur.

Preparation:
    .
   
Description:
Glyphs have been ordered according to the similarity of their probability distributions.
X-axis : Position of a letter within a word:: Leftmost -> beginning of word,  Rightmost -> end of word 
Y-axis : Probability::  bottom is 0 (never occurs) -> top is 1 (certain to occur)
Labels : EVA-letter (black), VMS Glyph (gray)
Low frequency glyphs not shown
Underlying grey plots are the exact plots
Colored overlays are the Loess smoothed data ( to reduce noise )

Explanation:
We can see with the group < p, q, c, s > that the plot starts near the top left of the graph and descends quickly as it moves to the right.
Showing these glyphs have a high probability of appearing at the start of a word and have a low Probability at the end of a word.

The roughly opposite effect is observed in the groups  < y, r, m, g > and < l > and < n >.
Their plot starts at the bottom left, indicating a low probability of these letters occurring at the beginning of the word.
The plots stay low, denoting their continued low chance of being found as we proceeed further into the word.
Then their plots rise steeply showing that these letters are more likely to be found at the end of a word.

Letters < e > and < i > have an single peak in the middle of the X-axis indicating they are most likely to be found in the middle of a word.

=====================================

A generalized grouping can be described like so:

For a detailed study on glyph position see < S.Palmer, Voynich MS glyph position stacks >
You are not allowed to view links. Register or Login to view.

N.B
3 Statistical artefacts are noted:
< EVA-q > the loess smooothing goes awry here, it should follow the black line better.
< EVA-o > the graph shows a small second peak at about 2/3 of the X-axis, this peak is not in the data source.
< EVA-l >, < EVA-n > are colored differently because of the statistic used to generate the colored groups.
    .

This is a very intuitive visualisation, Rob, well done! I might refer to this in one of my videos later, I think it would work well to get some ideas across.

Do you have a code for generating these? (I haven't been able to check the linked website yet, not on my own pc right now).

How does this compare with regular languages or relevant period texts? 

Interesting that EVA (d) is clearly a "two-location" glyph that stands apart from the others. Is this comparable in other text sources?

This is really nice work Rob thank you

I must admit that only today I checked up on Renes site at the EVA lists again because it struck me I didn't know what anyone had labelled "curvy n" or "straight d" as I'm guilty of using the Voynichese website maybe more than I should. 

It looks like "curvy n" (b) probably works like "n" does. Maybe "straight d" (j) has some influence on the multiple locations R.Sale points out. If I had to guess I'd say probably not, but I'll definitely keep an eye out as looking into d/j was on my list to do anyway.

Caveat: I barely know R-language so i struggled a bit to make this work, also the original used a Million words whereas this just uses the 37k ish voynich words however as far as i can tell, the resultant graphs agree with the raw data ( except for the small discrepancy with EVA-o ) and parallel S.Palmers graphs, so i think its all good but if anyone knows of any mistakes i've made here, please point them out.

@Koen, R-code is in the link, i had to change the 'datasource' bit to load and process the edited vms text.

@R.Sale, graphs for English are on the OG blog post, i copied and grouped them in an image but im unsure if i can post that image because of copyright.
As to how it compares with regular languages or relevant period texts, i have no idea.
This is vms EVA and Hargreaves did English , so they are the only 2 available for comparison that i know of.
Some quick observations:
English-C has a graph resembling EVA-d.
English <B,J,Q,W> resemble the  EVA-<p,q,c,s> group.
English-U is most similar to EVA-e.

i have no idea what that means or even if it means anything.
I just saw the OG blog post and thought that looks fun, lets try it with EVA and it may help me visualize Stolfis' core-crust-mantle idea.

For comparison, articles by Sean B. Palmer:

You are not allowed to view links. Register or Login to view.
Voynich MS glyph position stacks
These charts show where glyphs appear within words in the Voynich MS. Rows represent all words of a particular length, from words of length one at the top to nine at the bottom. Darker blue indicates more appearances in that position.

You are not allowed to view links. Register or Login to view.
Glyphs within Voynich MS words
This diagram shows the relative positions of glyphs within words in the Voynich MS. A blue square indicates that top comes after left, and a red square vice versa.

Wow - graphs I can understand!  Bookmarking…