Characterizing random variables in the context of signal detection theory

摘要: Abstract By generalizing properties of the standard normal model signal detection theory, we are able to characterize symmetric, partially stable random variables in terms a pair observable relations coupling yes-no and forced-choice paradigms. One these is provided by ‘area theorem’: area subtended receiver operating characteristics equals probability correct response corresponding task. The other relation involves algebraic notion bisymmetry. It turns out that most actually . Accordingly can interpret our results as characterizing models which ‘signal’ ‘noise’ arise sums many small independent contributions.