Worksheet effect
Fitting
This page shows an attempt to compensate for the
worksheet effect, using the
same method
as in compensating for the fieldworker effect.
data: raw phonetic strings, no lexical variants
subset: fieldworker: Lowman
worksheets: Middle Atlantic, South Atlantic
752 informants
grouping: fieldworker + worksheet + community: 361 locations
(two locations merged because of identical coordinates)
method: Levenshtein
triangulation: ignoring Preliminary South Atlantic worksheet
Mean and standard deviation of uncorrected differences. Note the much larger
differences between the two sets of differences than within each set of
differences:
The maps below show the result of the correction.
maps: stochastic clustering, group average + weighted average
The uncorrected map shows a complete trace of the worksheet border. Also note
that the differences in the South Atlantic are stronger than the
differences in the Middle Atlantic:
After correction, much of the trace of the worksheet border has disappeared.
Dialect borders in Middle and South now show equally strong:
Correction removes much of the worksheet border, and has only a limited effect
on other borders:
Using coloured clusters, you get a clearer picture of the global effect.
maps: stochastic clustering, group average + weighted average
followed by classical MDS on cophenetic differences
The worksheet effect is still there, visible in West Virginia, but after
correction, the effect is less strong:
