search

pymc-devs / pymc-examples

Examples of PyMC models, including a library of Jupyter notebooks.
https://www.pymc.io/projects/examples/en/latest/
MIT License
269 stars 234 forks source link

Revised multilevel modeling notebook #468

Closed ghost closed 1 year ago

ghost commented 1 year ago

This is an update to the multilevel modeling notebook that makes two main changes:

  1. Removes the independent floor vs basement parameterization from the current notebook. Expressing the floor and basement measurements as entirely independent is not a great starting point, given that they are measuring the exact same phenomenon, just from a different location. This reverts to the covariate model that simply uses floor as an offset. I think this is less confusing overall.
  2. Adds the non-centered parameterization. While perhaps redundant with another notebook, its an important component of modeling this problem correctly.

I am also removing the GLM-hierarchical notebook, as it is almost completely redundant with this one, now that it no longer uses the extinct glm submodule.

review-notebook-app[bot] commented 1 year ago

Check out this pull request on  ReviewNB

See visual diffs & provide feedback on Jupyter Notebooks.

Powered by ReviewNB

twiecki commented 1 year ago

So far I only skimmed but this looks excellent. Still need to solve pre-commit issues with the myst files.

fonnesbeck commented 1 year ago

Pre-commit issues fixed.

review-notebook-app[bot] commented 1 year ago

View / edit / reply to this conversation on ReviewNB

twiecki commented on 2022-11-25T14:46:17Z ----------------------------------------------------------------

Can we add axis labels?


fonnesbeck commented on 2022-11-25T22:56:19Z ----------------------------------------------------------------

Added.

review-notebook-app[bot] commented 1 year ago

View / edit / reply to this conversation on ReviewNB

twiecki commented on 2022-11-25T14:46:18Z ----------------------------------------------------------------

Curious why these look so different from before.


fonnesbeck commented on 2022-11-25T16:29:36Z ----------------------------------------------------------------

Probably a function of the odd parameterization that is in the current version (assuming floor and basement are independent). This revision matches what you see in Gelman.

review-notebook-app[bot] commented 1 year ago

View / edit / reply to this conversation on ReviewNB

twiecki commented on 2022-11-25T14:46:19Z ----------------------------------------------------------------

You could link to my blog post where I go into more detail: https://twiecki.io/blog/2017/02/08/bayesian-hierchical-non-centered/


fonnesbeck commented on 2022-11-25T22:58:56Z ----------------------------------------------------------------

Added a link.

review-notebook-app[bot] commented 1 year ago

View / edit / reply to this conversation on ReviewNB

twiecki commented on 2022-11-25T14:46:20Z ----------------------------------------------------------------

You could also mention somewhere that instead of a hierchical model, we can place geospatial distribution over the counties radon levels like here: https://www.pymc-labs.io/blog-posts/spatial-gaussian-process-01/


fonnesbeck commented on 2022-11-25T23:02:36Z ----------------------------------------------------------------

Added a link.

fonnesbeck commented 1 year ago

Probably a function of the odd parameterization that is in the current version (assuming floor and basement are independent). This revision matches what you see in Gelman.

View entire conversation on ReviewNB

fonnesbeck commented 1 year ago

Added.

View entire conversation on ReviewNB

fonnesbeck commented 1 year ago

Added a link.

View entire conversation on ReviewNB

fonnesbeck commented 1 year ago

Added a link.

View entire conversation on ReviewNB

twiecki commented 1 year ago

Great stuff, this classic is in great shape now!

  • © Githubissues.
  • Githubissues is a development platform for aggregating issues.