update model_comp examples and ppc_w to work with idata #4042
Conversation
|
Check out this pull request on Review Jupyter notebook visual diffs & provide feedback on notebooks. Powered by ReviewNB |
Codecov Report
@@ Coverage Diff @@
## master #4042 +/- ##
==========================================
+ Coverage 86.79% 86.81% +0.02%
==========================================
Files 88 88
Lines 14143 14150 +7
==========================================
+ Hits 12276 12285 +9
+ Misses 1867 1865 -2
|
There was a problem hiding this comment.
Thanks for this nice update @aloctavodia ! I spotted some typos below. Also, you'll need to merge master into this branch to avoid conflicts in the relase notes
RELEASE-NOTES.md
Outdated
| ### Maintenance | ||
|
|
||
| ### Documentation | ||
|
|
||
| ### New features | ||
| - `sample_posterior_predictive_w` can now feed on `xarray.Dataset` - e.g. from `InferenceData.posterior`. (see [#4042](https://github.com/pymc-devs/pymc3/pull/4042)) | ||
|
|
||
|
|
||
| ## PyMC3 3.9.3 (11 August 2020) | ||
|
|
There was a problem hiding this comment.
I updated the release notes in an earlier PR: can you merge master into this branch and then add a line for this PR please?
| @@ -1,25 +1,34 @@ | |||
| { | |||
There was a problem hiding this comment.
THisThis approximation is based on importance sampling- By default ArviZ, uses LOO, but WAIC is also available.
Reply via ReviewNB
| @@ -1,25 +1,34 @@ | |||
| { | |||
There was a problem hiding this comment.
Shouldn't you call ArviZ directly in these last two cells? Especially since you're talking about ArviZ in the text
Reply via ReviewNB
| @@ -1,25 +1,34 @@ | |||
| { | |||
There was a problem hiding this comment.
The first of this these functions is compare, which this one computes LOO (or WAIC) WAIC (or LOO) from a set of traces and models and returns a DataFrame.
Reply via ReviewNB
| @@ -1,25 +1,34 @@ | |||
| { | |||
There was a problem hiding this comment.
Maybe use Latex ordered list formatting for more beautiful display?
0) The index is are the names of the models taken from the keys of the dictionary passed to compare(.).
1) rank, the ranking of on the models starting from 0 (best model) to the number of models.
2) loo, the values of LOO (or WAIC). (needs a space after LOO)
5) weight, the weights assigned to each model. These weights can be loosely interpreted as the probability of each model being true (among the compared models) given the data
9)... Other options are deviance -- this is the log-score multiplied by -2 (this reverts the order: a lower higher LOO/WAIC will be better) -- and negative-log -- this is the log-score multiplied by -1 (as with the deviance scale, a lower value is better).
Reply via ReviewNB
| @@ -1,25 +1,34 @@ | |||
| { | |||
There was a problem hiding this comment.
The value of the highest LOO (i.e the best estimated model) is also indicated ...
Reply via ReviewNB
| @@ -1,25 +1,34 @@ | |||
| { | |||
There was a problem hiding this comment.
...there is little to choose between the models in this case, giving given that both models gives very similar values of the information criteria.
Reply via ReviewNB
No description provided.