Niche as a Determinant of Word Fate in Online Groups
Figure 5
Frequency as a predictor of falling below threshold and of frequency decay.
This figure is the -counterpart of Figure 3. The dashed green lines in panels B, E indicate the minimum possible
for a given
, due to the threshold
imposed at
. The analysis in Table 1 includes only the range
, where
and
are the limits of the range considered. The range is truncated at
because, for words above this frequency,
is so large compared to the number of users or threads that
is not informative. The range is truncated at
for comp.os.linux.misc (
for rec.music.hip-hop) because below these cutoffs the exclusion of words falling under the threshold (i.e.,
) introduces artifacts in the relationship to
(c.f. the relationship of the dashed green lines to the 10th percentile line). Specifically,
was chosen for each dataset so that the percentage of words falling below the threshold at
would be less than
of the words with
.