Open mitch-at-orika opened 7 months ago
Check out this pull request on
See visual diffs & provide feedback on Jupyter Notebooks.
Powered by ReviewNB
View / edit / reply to this conversation on ReviewNB
NathanielF commented on 2023-12-18T18:37:48Z ----------------------------------------------------------------
typo - "When [constants] are included
Also, when introducing this clarifying distinction it might be worth saying how this is occasionally mistaken for the need to 0 out the entire equation for the outside good, but the actual identification requirements are not as strict see reference etc...
View / edit / reply to this conversation on ReviewNB
NathanielF commented on 2023-12-18T18:37:49Z ----------------------------------------------------------------
Line #18. beta_ic * wide_heating_df["ic.hp"] + beta_oc * wide_heating_df["oc.hp"]
If you're changing this line you should remove of clarify the np.zeros line
View / edit / reply to this conversation on ReviewNB
NathanielF commented on 2023-12-18T18:37:50Z ----------------------------------------------------------------
Line #20. 0 + beta_ic * wide_heating_df["ic.hp"] + beta_oc * wide_heating_df["oc.hp"]
This is fine. I think but we should keep to this pattern in subsequent models then too.
View / edit / reply to this conversation on ReviewNB
NathanielF commented on 2023-12-18T18:37:51Z ----------------------------------------------------------------
Great! Both coefficients are negative. Immediately better and more interpretable out of the gate. More effective sample sizes too.
View / edit / reply to this conversation on ReviewNB
NathanielF commented on 2023-12-18T18:37:52Z ----------------------------------------------------------------
Interesting that there appears to be wider uncertainty now. I guess we have more free parameters and they just propagate. I think the right interpretation is to stress a bit more on the freedom to specify individual utility for goods that is unlocked in this approach to the models
View / edit / reply to this conversation on ReviewNB
NathanielF commented on 2023-12-18T18:37:52Z ----------------------------------------------------------------
Check that this specification makes sense after re-fit.
View / edit / reply to this conversation on ReviewNB
NathanielF commented on 2023-12-18T18:37:53Z ----------------------------------------------------------------
Maybe add a note under these model comparison plots about how the freedom of the utility specification drives our understanding about the various components of market demand, and how the bayesian model development workflow helps uncover and validate the influences on desire
View / edit / reply to this conversation on ReviewNB
NathanielF commented on 2023-12-18T18:37:54Z ----------------------------------------------------------------
I think we should change this model too, but there is an argument that it's worth show-casing the common "zero-out" method too. But if you leave it as is you should add a note to explain the different approach.
View / edit / reply to this conversation on ReviewNB
NathanielF commented on 2023-12-18T18:37:55Z ----------------------------------------------------------------
Not necessary but it might be nice to overlay the 50% HDI on top of the 95% hdi you see here. It's more Bayesian in spirit and shows the centre of the distributions better.
Thanks for the PR @mitch-at-orika! Sorry it took me so long to get back to you - Christmas has been waaay too busy. I've added a few comments above. Mostly about deciding how consistent we want to be in updating the model specification and making sure the rest of the text reflects the new model specification.
On the pre-commit checks it looks like you just need to reset the notebook and run all cells from the top.
Testing one alternative utility not zero (just the constant)
Issue #617
:books: Documentation preview :books:: https://pymc-examples--618.org.readthedocs.build/en/618/