Skip to content

Docs: non-stationarity — likelihood tempering vs time-varying parameters #2517

@daimon-pymclabs

Description

@daimon-pymclabs

The existing time-varying parameters notebooks (Time-Varying Media, Time-Varying Parameters) show how to use the feature but don't explain why you'd choose it over other approaches to non-stationarity, such as likelihood tempering/weighting.

A documentation page (either standalone or as an expanded introduction to an existing TVP notebook) covering the conceptual distinction would be valuable: adjusting inference (reweighting the likelihood to favour recent data) vs modelling the changing data-generating process directly (time-varying parameters). This framing came up in an EAP advisory discussion where a production MMM team had implemented likelihood tempering and needed guidance on the tradeoffs.

[@drbenvincent suggested this]

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions