The effects on areal estimates obtained by aggregating class predictions for individual pixels are deviation of the estimates from their true values, erroneous confidence intervals, and incorrect inferences.
We evaluated the relative importance of each source of uncertainty in determining the difference between the least-cost paths (LCPs) costs and resistance distances generated for the true and erroneous cost-surfaces, using four model evaluation metrics.
Moreover, precise inference is more difficult due to incomplete annotation, along with different stringency criteria, customary thresholds to classify true and erroneous AS events, and various gene models used in different species.
This was especially true when the erroneous examples were similar to the kinds of errors that they had committed [14].
(Nonetheless, unless we are fundamentally confused about something, all our beliefs meet this formal condition.[3]) But there is a further question: which of our beliefs are actually (or "materially") true, and which erroneous?
True errors, such as erroneous reactions, can produce misleading results.
Noise in source data can push computed results away from their true values, lead to erroneous reporting.
Although the benefits of using such technology are clear, variants can be called that are interpreted as being true variants but are erroneous [ 15].
With signal and noise of equal size for example, in ~50% of cases (Fig. 6) the optimum detected will not be the true optimum but an erroneous alternative.
Furthermore, there are very few computational methods available to distinguish true viral variants from erroneous mutations, with the majority of methods tailored to the 454 platform such as AmpliconNoise [ 29] and ReadClean454 [ 30].
