't' and 'k' are surely the most similar characters for what regards the distribution of the preceding and of the following character. The geometrical distances between the distributions of 'k' and 't' are 0.11 (previous character) and 0.13 (following character) [min possible value = 0, max possible value = 1.414]. Notice this does not consider JoJo's post You are not allowed to view links. Register or Login to view. (there the discrepancies are beyond the immediately following character).
'ch' and 'sh' are rather similar too: 0.18 (previous character) and 0.13 (following character). 'cth' and 'cfh' get 0.16 (previous) and 0.13 (following), while 'ckh' and 'cth' have 0.34 (previous, rarther different) and 0.18 (following).
The remaining 'similar' pairs are almost always only for the following character, and they come from the characters which overwhelmingly sit at the end of words (ie. both 'n' and 'm' are mostly followed by 'space': the distance is 0.03 for the following character, but a whopping 1.18 for the previous character).
As an additional observation, 'ok' and 'ot' are even more similar than 'k' and 't' for what regards the following character (distance = 0.11), but rather distant for the previous (0.31)
PS.: all the statistics calculated on the whole RF1a-n transcroption.
'ch' and 'sh' are rather similar too: 0.18 (previous character) and 0.13 (following character). 'cth' and 'cfh' get 0.16 (previous) and 0.13 (following), while 'ckh' and 'cth' have 0.34 (previous, rarther different) and 0.18 (following).
The remaining 'similar' pairs are almost always only for the following character, and they come from the characters which overwhelmingly sit at the end of words (ie. both 'n' and 'm' are mostly followed by 'space': the distance is 0.03 for the following character, but a whopping 1.18 for the previous character).
As an additional observation, 'ok' and 'ot' are even more similar than 'k' and 't' for what regards the following character (distance = 0.11), but rather distant for the previous (0.31)
PS.: all the statistics calculated on the whole RF1a-n transcroption.