Merge remote-tracking branch 'upstream/main'

Author: fonnesbeck

.github/workflows/pre-commit.yml

@@ -15,5 +15,5 @@ jobs:
     - uses: actions/setup-python@v2
     - uses: actions/setup-node@v2
       with:
-        node-version: '15'
+        node-version: '18'
     - uses: pre-commit/[email protected]

.github/workflows/rtd-link-preview.yml

@@ -0,0 +1,16 @@
+name: Read the Docs Pull Request Preview
+on:
+  pull_request_target:
+    types:
+      - opened
+
+permissions:
+  pull-requests: write
+
+jobs:
+  documentation-links:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: readthedocs/actions/preview@v1
+        with:
+          project-slug: "pymc-examples"

.jupytext.toml

@@ -1,2 +1,2 @@
-notebook_metadata_filter = "substitutions,-jupytext.text_representation.jupytext_version"
+notebook_metadata_filter = "myst,-jupytext.text_representation.jupytext_version"
 formats = ["ipynb", ".myst.md:myst"]

.pre-commit-config.yaml

@@ -1,26 +1,25 @@
 repos:
 - repo: https://github.com/psf/black
-  rev: 22.3.0
+  rev: 23.7.0
   hooks:
     - id: black-jupyter
 - repo: https://github.com/nbQA-dev/nbQA
-  rev: 1.1.0
+  rev: 1.7.0
   hooks:
     - id: nbqa-isort
       additional_dependencies: [isort==5.6.4]
     - id: nbqa-pyupgrade
       additional_dependencies: [pyupgrade==2.7.4]
       args: [--py37-plus]
 - repo: https://github.com/MarcoGorelli/madforhooks
-  rev: 0.3.0
+  rev: 0.4.1
   hooks:
     - id: check-execution-order
       args: [--strict]
       exclude: |
             (?x)^
             ^examples/ode_models/ODE_with_manual_gradients\.ipynb$
             |examples/samplers/DEMetropolisZ_EfficiencyComparison\.ipynb$
-            |examples/gaussian_processes/GP-Latent\.ipynb$
             |examples/gaussian_processes/GP-MaunaLoa2\.ipynb$
             |examples/samplers/MLDA_gravity_surveying\.ipynb$
             |examples/howto/sampling_callback\.ipynb$
@@ -33,7 +32,7 @@ repos:
             |examples/samplers/MLDA_variance_reduction_linear_regression\.ipynb$
 
 - repo: https://github.com/FlamingTempura/bibtex-tidy
-  rev: v1.8.5
+  rev: v1.11.0
   hooks:
     - id: bibtex-tidy
       files: examples/references.bib
@@ -97,7 +96,7 @@ repos:
       language: pygrep
       types_or: [markdown, rst, jupyter]
 - repo: https://github.com/mwouts/jupytext
-  rev: v1.13.7
+  rev: v1.15.1
   hooks:
   - id: jupytext
     files: ^examples/.+\.ipynb$

examples/case_studies/bart_heteroscedasticity.ipynb renamed to examples/bart/bart_heteroscedasticity.ipynb

@@ -24,8 +24,6 @@ kernelspec:
 In this notebook we show how to use BART to model heteroscedasticity as described in Section 4.1 of [`pymc-bart`](https://github.com/pymc-devs/pymc-bart)'s paper {cite:p}`quiroga2022bart`. We use the `marketing` data set provided by the R package `datarium` {cite:p}`kassambara2019datarium`. The idea is to model a marketing channel contribution to sales as a function of budget.
 
 ```{code-cell} ipython3
-:tags: []
-
 import os
 
 import arviz as az
@@ -37,8 +35,6 @@ import pymc_bart as pmb
 ```
 
 ```{code-cell} ipython3
-:tags: []
-
 %config InlineBackend.figure_format = "retina"
 az.style.use("arviz-darkgrid")
 plt.rcParams["figure.figsize"] = [10, 6]
@@ -157,8 +153,6 @@ The fit looks good! In fact, we see that the mean and variance increase as a fun
 ## Watermark
 
 ```{code-cell} ipython3
-:tags: []
-
 %load_ext watermark
 %watermark -n -u -v -iv -w -p pytensor
 ```

examples/case_studies/bart_heteroscedasticity.myst.md renamed to examples/bart/bart_heteroscedasticity.myst.md

@@ -1037,7 +1037,7 @@
    "id": "2b680d91",
    "metadata": {},
    "source": [
-    "This plot helps us understand the season behind the bad performance on the test set: Recall that in the variable importance ranking from the initial model we saw that `hour` was the most important predictor. On the other hand, our training data just sees `hour` values until $19$ (since is our train-test threshold). As BART learns how to partition the (training) data, it can not differentiate between `hour` values between $20$ and $22$ for example. It just cares that both values are greater that $19$. This is very important to understand when using BART! This explains why one should not use BART for time series forecasting if there is a trend component. In this case it is better to detrend the data first, model the remainder with BART and model the trend with a different model."
+    "This plot helps us understand the reason behind the bad performance on the test set: Recall that in the variable importance ranking from the initial model we saw that `hour` was the most important predictor. On the other hand, our training data just sees `hour` values until $19$ (since is our train-test threshold). As BART learns how to partition the (training) data, it can not differentiate between `hour` values between $20$ and $22$ for example. It just cares that both values are greater that $19$. This is very important to understand when using BART! This explains why one should not use BART for time series forecasting if there is a trend component. In this case it is better to detrend the data first, model the remainder with BART and model the trend with a different model."
    ]
   },
   {

examples/case_studies/BART_introduction.ipynb renamed to examples/bart/bart_introduction.ipynb

@@ -389,7 +389,7 @@ ax.set(
 );
 ```
 
-This plot helps us understand the season behind the bad performance on the test set: Recall that in the variable importance ranking from the initial model we saw that `hour` was the most important predictor. On the other hand, our training data just sees `hour` values until $19$ (since is our train-test threshold). As BART learns how to partition the (training) data, it can not differentiate between `hour` values between $20$ and $22$ for example. It just cares that both values are greater that $19$. This is very important to understand when using BART! This explains why one should not use BART for time series forecasting if there is a trend component. In this case it is better to detrend the data first, model the remainder with BART and model the trend with a different model.
+This plot helps us understand the reason behind the bad performance on the test set: Recall that in the variable importance ranking from the initial model we saw that `hour` was the most important predictor. On the other hand, our training data just sees `hour` values until $19$ (since is our train-test threshold). As BART learns how to partition the (training) data, it can not differentiate between `hour` values between $20$ and $22$ for example. It just cares that both values are greater that $19$. This is very important to understand when using BART! This explains why one should not use BART for time series forecasting if there is a trend component. In this case it is better to detrend the data first, model the remainder with BART and model the trend with a different model.
 
 +++