deserialize_alternative_prior#

pymc_marketing.special_priors.deserialize_alternative_prior(data)[source]#

Alternative deserializer that recursively handles all nested parameters.

This handles the flat {"distribution": ...} prior format used by the MMM YAML schema, where any parameter might itself be a nested prior, and also extracts the centered and transform parameters.

Examples

This handles cases like:

distribution: Gamma
alpha: 1
beta:
    distribution: HalfNormal
    sigma: 1
    dims: channel
dims: [brand, channel]