We improve the performance to work with more registers in vector files. For
example, over two hundred thousand polygons in the map under assessment.
The difference between both plugins is that Accuracy Assessment only process the
raw matrix and does not take into account the sampling design used to select the
reference data.
AccurAssess carries out a correction of bias of sampling before computing the
accuracy indices (Method of Card 1982).
In brief, Accuracy Assessment will only give correct indices if the number of
reference sample by class is proportional to the area of each class on the map
(e.g. simple random or systematic sampling but not in case of stratified
sampling except the rare case of a stratification which respects these
proportions). AccurAssess correct the bias of sampling and will give correct
results with simple random and stratified sampling.
Additionally, AccurAssess give the confidence intervals (CI) around the
estimates of accuracy indices and gives corrected estimates of class area (also
with CI).
see the paper for more details:
http://www.hindawi.com/journals/geography/2014/372349/