Skip to content

Expand docs: more info required about aggregating the causal impact over time #498

Open
@drbenvincent

Description

@drbenvincent

For the interrupted time series docs here https://causalpy.readthedocs.io/en/stable/notebooks/its_pymc.html we have an example causal impact over time.

Image

The docs then talk about aggregating these values over time (mean), reporting that value, but also giving a warning.

Image

A good point was raised in the pymc discourse which indicates more information is needed here.

This issue could be closed by:

  • expanding the discussion of reducing the causal impact over the whole post-intervention period into a single distribution.
  • Talking about when and why different aggregation functions could be useful:
    • mean - if you have a step change in the post-intervention period. Would give you the mean increase per time period
    • sum - gives you the total effect. Note that this is equivalent to looking at the final point in the cumulative causal impact.

Metadata

Metadata

Assignees

No one assigned

    Labels

    documentationImprovements or additions to documentation

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions