6.3. Probability distributions
Probability distributions are important when probability is used
to model uncertainty. The validity of risk estimates relies on the
assumptions that selected probability distributions are valid models
of actual randomness. This is not always the case, which is illustrated
by recent examples from economics (Taleb, 2004).
Another complication is that random processes are not necessarily
stationary, which means that available observation time
may not be enough for obtaining reliable estimates of the probability
distributions involved. Even if the source of randomness in a
system can be assumed to be white noise, the signal may go
through filters and there may be interactions through feedback
and feed forward loops that influence probability distributions.