A function to optimize MSTTR segment sizes
This function calculates an optimized segment size for MSTTR.
segment.optimizer(txtlgth, segment = 100, range = 20, favour.min = TRUE)
txtlgth |
Integer value, size of text in tokens. |
segment |
Integer value, start value of the segment size. |
range |
Integer value,
range around |
favour.min |
Logical, whether as a last ressort smaller or larger segment sizes should be prefered, if in doubt. |
When calculating the mean segmental type-token ratio (MSTTR), tokens are divided into segments of a given size and analyzed. If at the end text is left over which won't fill another full segment, it is discarded, i.e. information is lost. For interpretation it is debatable which is worse: Dropping more or less actual token material, or variance in segment size between analyzed texts. If you'd prefer the latter, this function might prove helpful.
Starting with a given text length, segment size and range to investigate,
segment.optimizer
iterates through possible segment values. It returns the segment size which would drop the fewest
tokens (zero, if you're lucky). Should more than one value fulfill this demand,
the one nearest to
the segment start value is taken. In cases,
where still two values are equally far away from the
start value,
it depends on the setting of favour.min if the smaller or larger segment size
is returned.
A numeric vector with two elements:
seg |
The optimized segment size |
drop |
The number of tokens that would be dropped using this segment size |
segment.optimizer(2014, favour.min=FALSE)
Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.