Support for categorical/grouping variable to enable varying intercepts/slopes and interactions

[LacombeLouis]

Allow a single model workflow to incorporate a categorical/grouping variable so that the model can estimate category-specific effects, for example varying intercepts and/or varying slopes on media and control variables without manually building many separate models or hand-crafting large numbers of interaction columns.

What is the recommended / supported approach for doing this in Meridian?

• native support for specifying a grouping factor that automatically produces varying intercepts and slopes (partial pooling / hierarchical priors) and how to express that in configuration?
• guidance on priors / pooling strategies when categories have uneven data or many levels, and performance/scalability considerations for many categories?

[cpulavarthi]

Hello @LacombeLouis,

Thank you for contacting us!

It looks like your requirements can be met by using a different categorical variable other than geo in the hierarchical modeling framework to help you estimate that variable-specific effects.

Meridian's hierarchical modeling structure was designed for geo-level modeling, which has important advantages. Although it is possible to use another categorical variable in place of geographic region as the hierarchical variable, this is generally not recommended. If another hierarchical variable is to be used, it must be possible to stratify both the KPI and the media units by this variable.
You may check the FAQ page that talks about non-geo hierarchies in more detail.

Feel free to reach out if you have any further questions or suggestions for us!

Google Meridian Support Team