The log posterior predictive can be computed using:
p = logpostpred(d::ConjugatePostDistribution, x)
resulting in a scalar if x
and d
is univariate or if x
and d
is multivariate. Otherwise, a vector is returned.
Observations can be added to the conjugate posterior distribution using:
add!(d::ConjugatePostDistribution, x)
for inplace operations or using
d = add(d::ConjugatePostDistribution, x)
resulting in a new distribution d
.
Observation can be removed from a conjugate posterior distribution using:
remove!(d::ConjugatePostDistribution, x)
for inplace operations of using
d = remove(d::ConjugatePostDistribution, x)
resulting in a new distribution d
.
The posterior parameters of a distribution can be obtained using:
parameters = posteriorParameters(d::ConjugatePostDistribution)
those parameters are returned as tuples, e.g. parameters = (μ, σ)
in the case of NormalNormal distributions.