
Metrics for Decision Support ◾ 311
defects and the less complex the problem the fewer defects. Finally, how
many defects you nd is inuenced not just by the number there to nd
but also by the amount of testing eort.
e task of dening the NPTs for each node in this model is clearly more
challenging than the previous examples. In many situations we would need
the NPT to be dened as a function rather than as an exhaustive table of
all potential parent state combinations. Some of these functions are deter-
ministic rather than probabilistic: for example, the “Residual defects” is
simply the numerical dierence between the “Defects inser ...