Lightweight Mmm Versions Save

Minor updates:

• bugfix in plot_response_curves() function
• change numpy version in requirements

What's Changed

Minor bugfix:

• Normalise the scale of the media priors for a hierarchical (geo-level) model so that the mean of the priors is the same as in the non-hierarchical (national-level) model.

minor update to help resolve dependency issues

What's Changed

• Deprecated Python 3.7 support and updated requirements.
• New data quality analysis added to example Colabs.
• Internal changes in preparation for a larger update.
• Minor fixes and improvements.

What's Changed

• Fixed optimization previous media/budget allocation to account for unequal prices.
• Exposed more scipy's parameters on optimization and made jacobian more accurate.
• New plot for prior-posterior comparison for all parameters in the model.
• Minor fixes and improvements.

What's Changed

• Add new API for custom priors and respective documentation.
• Further expansion on supporting python version from 3.7 to 3.10 inclusive.
• Documentation improvements.
• Minor fixes.

What's Changed

• Plotting media contribution overtime has been added to plots.
• Plotting results from optimisation has been added.
• New utility for going from pd.DataFrame to our expected geo 3D array for geo models
• Expanded documentation of models.
• Improved workflows, CI and documentation
• Fixed edge case for certain tests
• Minor changes and fixes

Full Changelog: https://github.com/google/lightweight_mmm/compare/0.1.3...v0.1.4

What's Changed

• Improved workflows and documentation
• Fixed edge case for hill function
• Minor changes and fixes

What's Changed

• Add geo level models for bigger countries
• Adapt data scaling to accept geo data
• Adapt media optimization to geo models
• Adapt plots to geo level model and data
• Minor fixes and changes

What's Changed

• Fix install and import issues
• Add simple demo notebook
• Add control over seeds in fit and predict methods
• Minor changes and fixes