bayesplain.priors.ConcentrationPrior

class bayesplain.priors.ConcentrationPrior(a, name='custom', rationale='')[source]

A symmetric Dirichlet prior over the cells of a contingency table.

The same preset names as BetaPrior, resolving to the concentration a that Gunel and Dickey’s Bayes factors take. The correspondence is exact rather than analogous: for a 2x2 table built from successes and failures, concentration a on the columns is a Beta(a, a) prior on each group’s rate.

Larger values pull the alternative’s predictions toward the null. Jamil et al. (2017) put it well: a = 10 behaves like a flat a = 1 prior that has already been updated with nine hypothetical observations in every cell.

Variables:
a : float

The concentration.

name : str

Short label used in printed output.

rationale : str

Plain-English statement of what this assumption commits you to.

Methods

dist()

Return None: a table prior has no one-dimensional density.

a[source]
dist()[source]

Return None: a table prior has no one-dimensional density.

The prior lives on the simplex of cell probabilities, so there is no single curve to overlay on a posterior plot the way there is for a rate. Plotting code checks for None and skips the prior layer.

property label[source]

Name and parameter together, for printing in a summary table.

name = 'custom'[source]
rationale = ''[source]