Correlation between line-initial letter and words on that line?

[obelus]

nickpelling --

Attached are visualizations of the line-initial-character-specific word frequencies in an FSG text sample (as specified below).

The rows are indexed by line-initial character, together with the total number of pooled words for each;  the columns are indexed by words that appear in the initial-character-specific pools.  Both axes are ordered by frequency in the complete text sample, with greatest frequency at the upper left, and the 40 most-frequent words shown.

This grayscale indicates absolute frequency of a word within each initial-character word pool:

   

Here, the previous values are normalized by the frequency of each word in the complete text sample
(frequency in initial-letter pool) ÷ (global frequency)
on a logarithmically compressed color scale;  unity is beige, and saturated hues are deviant:

   

It appears as if the popular line-initial letters (O, G, 8, etc), with their concomitant large pools of words, independently approach the global word-frequency distribution -- suggestive of a null result for initial-letter significance. At least some of the deviant distributions for rare initial letters must be statistical fluctuations.

TEXT SAMPLE:  FSG_ivtff_1c.txt extracted with ivtt options -c5 -p2 -s1 -l4 -w1 -h2 -u1 +@P
("normal" running paragraph text, with preferred readings of alternatives, and uncertain spaces and questionable glyphs removed)

EDIT:  EVA formatting removed from secondary axes