Currently, nearly all rate parameters are defined as RealScalar<PositiveReal>. This makes it impossible to use a rate of 0 or to apply any prior defined on RealScalar<NonNegativeReal> (e.g. the Exponential prior).
Is this the intended behavior, or would it make sense to use the wildcard type RealScalar<? extends NonNegativeReal>?
(On a related note: Gamma distributions can be applied, but only because their domain is (incorrectly?) set to PositiveReal.)
Currently, nearly all rate parameters are defined as
RealScalar<PositiveReal>. This makes it impossible to use a rate of 0 or to apply any prior defined onRealScalar<NonNegativeReal>(e.g. theExponentialprior).Is this the intended behavior, or would it make sense to use the wildcard type
RealScalar<? extends NonNegativeReal>?(On a related note: Gamma distributions can be applied, but only because their domain is (incorrectly?) set to
PositiveReal.)