LinkSpec

class pymc_marketing.mmm.link.LinkSpec

Strategy object that centralises all link-dependent behaviour.

Subclasses implement the five link-specific decisions:

• inverse_link() – map the linear predictor to the response scale.

• default_likelihood() – default likelihood prior.

• default_intercept() – default intercept prior.

• validate_target() – fit-time target checks.

• create_media_contribution_deterministic() – graph for total_media_contribution_original_scale.

One concrete helper is shared by all links: create_total_response_deterministic() (the mu-effect objective total_response_original_scale, registered by MMM.build_model only when the model has mu effects).

Methods

LinkSpec.__init__(*args, **kwargs)
LinkSpec.create_media_contribution_deterministic(...)	Register total media contribution deterministic nodes.
LinkSpec.create_total_response_deterministic(...)	Register total_response_original_scale.
LinkSpec.default_intercept(dims)	Return the default intercept prior for this link.
LinkSpec.default_likelihood(dims)	Return the default likelihood prior for this link.
LinkSpec.inverse_link(mu)	Map the linear predictor mu to the response scale.
LinkSpec.mean_correction(posterior[, output_var])	Per-draw factor converting median-scale outputs to the response mean.
LinkSpec.original_scale_transform(variable, ...)	Transform a model variable to the original (response) scale.
LinkSpec.validate_likelihood_compatibility(...)	Raise if likelihood is incompatible with link.
LinkSpec.validate_target(y)	Validate that y is compatible with this link function.

Attributes

link