Bayes Window¶
Pretty and easy hierarchical Bayesian or linear mixed effects estimation with data overlay
Free software: MIT license
Documentation: https://mmyros.github.io/bayes-book/index.html
INSTALLATION¶
For gpu:
`pip install --upgrade "jax[cuda102]" -f https://storage.googleapis.com/jax-releases/jax_releases.html`
or cudaXXX from nvcc –version
Features¶
See extensive examples https://mmyros.github.io/bayes-book/index.html
TODO¶
Fit checks for LME under plot_diagnostics()
Posterior predictive to mimic data plot
Less haphazard testing
Shrinkage layer: Should be as easy as alt.layer with unpooled model
Prior predictive and posterior predictive plots
wait for resolution to https://github.com/vega/vega-lite/issues/4373#issuecomment-447726094
Random-effect plots (eg intercepts): bw.plot_extras?
Decide on Vega theme