Set custom priors to hill parameters

[cetagostini-wise]

Hello Google team,

It's possible to specify custom priors in the hill parameters directly? I saw, spec.ModelSpec(paid_media_prior_type='coefficient') but not sure if this is what I'm looking for. I would like to specify the parameters around shape and slope parameters for the hill, maybe it's possible but I didn't dig to deep in the documentation.

Thanks for the help!

[cpulavarthi]

Hello @cetagostini-wise,

Thank you for contacting us!

You have correctly identified that the coefficient priors should be used to customize priors for hill parameters. Prior distributions for slope and ec parameters can be set for media, reach & frequency, organic media and organic R&F channels. Please check our API Documentation on PriorDistribution for more details regarding this. The default distributions assumed by Meridian for these parameters can also be found in our documentation on Default Priors.

Section 2.2 of the research paper on Bayesian Methods for Media Mix Modeling with Carryover and
Shape Effects discusses how the Hill function is used to model shape effects used in Meridian. You may go through that for a better understanding to guide your prior selection and modeling choices.

Feel free to reach out for any further questions regarding this.

Thank you

Google Meridian Support Team