R/model-averaging-plots-posterior-helpers.R
simplify_as_mixed_posterior_bias.RdExtracts the posterior of one component of a bias mixture from
model-averaged posterior samples that hold the whole mixture as one
bias element, so that the component can be handled like any other
scalar parameter (for example plotted or summarized as a single
parameter). The component keeps the model-averaged posterior over all bias
branches: in the prior_list of the result, the branches without the
component are replaced by a null prior (a point at zero for PET and PEESE,
prior_none() for omega) that keeps their model weights.
simplify_as_mixed_posterior_bias(samples, parameter)the output of as_mixed_posteriors() or mix_posteriors():
a list with an element bias of class mixed_posteriors.bias.
the component to extract: "PET", "PEESE", or
"omega".
simplify_as_mixed_posterior_bias() returns samples
without the bias element and with the element
samples[[parameter]]: the mixed posterior of the component (a single
column for "PET" and "PEESE", the omega columns for
"omega", the latter with the class mixed_posteriors.weightfunction)
and its metadata. The other elements of samples are unchanged.
samples is the output of as_mixed_posteriors() or
mix_posteriors() for a bias mixture (e.g., a mixture of prior_PET(),
prior_PEESE(), prior_weightfunction() and prior_none() priors), whose
bias element has the class mixed_posteriors.bias. For
"PET" and "PEESE" the scalar column is the one of the bias
element; a bias mixture without that column is represented by the point
locations of its branches, which requires every active branch to be a point
prior. The draw metadata (atoms, support, prior densities, quantities, and
measure refusals) and the prior_list of the component are subset to
it; the model weights of the branches are kept.