bayesian-workflow

Name: bayesian-workflow
Author: Learning-Bayesian-Statistics

153stars15forks

How to install bayesian-workflow?

learning-bayesian-statistics/baygent-skills/bayesian-workflow

$npx -y skills add Learning-Bayesian-Statistics/baygent-skills --skill bayesian-workflow

Installs into the current project.

Summary

Opinionated Bayesian modeling workflow with PyMC and ArviZ. Contains critical guardrails (nutpie sampler, prior/posterior predictive checks, LOO-PIT calibration, prior sensitivity checks, 94% HDI, non-centered parameterizations, reproducible seeds) that agents won't apply unpromp

Files · 1

View on GitHub

SKILL.md

1	# Bayesian Workflow
2
3	## Workflow overview
4
5	Every Bayesian analysis follows this sequence. Do not skip steps -- especially model criticism.
6
7	1. Formulate — Define the generative story. What underlying process, that we're precisely trying to model, created the data?
8	2. Specify priors — See [references/priors.md](references/priors.md)
9	3. Implement in PyMC — Write the model. Prefer PyMC 5+ syntax. Use the latest version possible.
10	4. Run prior predictive checks — `pm.sample_prior_predictive()`. Verify priors produce plausible data ranges before fitting
11	5. Inference — `pm.sample(nuts_sampler="nutpie")`. Always use nutpie for speed (the nutpie python package provides cutting-edge sampling). Don't hardcode the number of chains — let the sampler pick the best default for the platform.
12	6. Diagnose convergence — Use `arviz_stats.diagnose(idata)` as the first check (requires arviz-stats >= 1.0.0). It covers R-hat, ESS, divergences, tree depth, and E-BFMI in one call. See [references/diagnostics.md](references/diagnostics.md)
13	7. Criticize the model — See [references/model-criticism.md](references/model-criticism.md)
14	8. Check prior sensitivity — Run `psense_summary(idata)` to verify conclusions are robust to prior choices. Visualize with `plot_psense_dist(idata)` from `arviz_plots`. Requires `log_likelihood` and `log_prior` in the InferenceData — compute them after sampling if needed. See [references/sensitivity.md](references/sensitivity.md)
15	9. Compare models (if applicable) — See [references/model-comparison.md](references/model-comparison.md)
16	10. Report results — Generate `<slug>/report.md` using the canonical template in [references/reporting.md](references/reporting.md). Run `scripts/check_diagnostics.py` to turn raw diagnostics into qualitative ratings + an ordered next-steps list, and use that output to fill the Assessment lines and Suggested Next Steps section. When the user mentions a non-technical audience or is new to Bayesian stats, additionally adapt the prose to plain language and include a glossary — but keep the canonical report structure as the audit trail.
17
18	## Installation
19
20	Prefer conda-forge / mamba-forge to install PyMC and its dependencies — pip can cause issues with
21	compiled backends (nutpie, JAX). Example:
22
23	```bash
24	mamba install -c conda-forge pymc nutpie arviz arviz-stats preliz
25	```
26
27	See Stack compatibility below for the PyMC 5.x vs 6.x / ArviZ 0.23 vs 1.x notes.
28
29	## Stack compatibility (PyMC 5.x and 6.x)
30
31	This skill teaches the latest PyMC 6 / ArviZ 1.x idioms and stays runnable on
32	PyMC 5.x during the transition (regulated/corporate environments can't always
33	upgrade freely). The scripts are verified on both stacks.
34
35	Most code is identical across versions. Where an API genuinely diverges, prefer the
36	form that runs on both:
37
38	\| Task \| Modern (PyMC 6 / ArviZ 1.x) \| Legacy (PyMC 5 / ArviZ 0.23) \|
39	\|---\|---\|---\|
40	\| Posterior-predictive plot \| `arviz_plots.plot_ppc_dist(idata)` — runs on both \| `az.plot_ppc(idata)` (removed in ArviZ 1.x) \|
41	\| Trace / rank plot \| `az.plot_trace(idata, var_names=[...])`; rank view `az.plot_rank(idata, var_names=[...])` — both run on both, but pass `var_names`: ArviZ 1.x errors when the auto-selected set exceeds its subplot cap (e.g. a vector `Deterministic` like `mu` with an `obs` dim) \| `az.plot_trace(idata, kind="rank_vlines")` (the `kind=` arg is 0.23-only) \|
42	\| Prior-predictive draws \| `pm.sample_prior_predictive(draws=500)` — runs on both \| `samples=500` (removed in PyMC 6) \|
43	\| Summary interval \| `az.summary(idata, ci_prob=0.94, ci_kind="hdi")` (ArviZ 1.x only) \| `az.summary(idata, hdi_prob=0.94)` (ArviZ 0.23 only) \|
44	\| Sampler output type \| `DataTree` \| `InferenceData` \|
45
46	`arviz_plots` (the ArviZ 1.x plotting package, imported as `azp`) installs on both
47	stacks, so leading with `azp.*` plots is the most portable choice. Ba

People who liked this also viewed

6 picks

Type

bayesian-workflow

153stars15forks

View on GitHub

How to install bayesian-workflow?

learning-bayesian-statistics/baygent-skills/bayesian-workflow

$npx -y skills add Learning-Bayesian-Statistics/baygent-skills --skill bayesian-workflow

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Bayesian Workflow
2
3	## Workflow overview
4
5	Every Bayesian analysis follows this sequence. Do not skip steps -- especially model criticism.
6
7	1. Formulate — Define the generative story. What underlying process, that we're precisely trying to model, created the data?
8	2. Specify priors — See [references/priors.md](references/priors.md)
9	3. Implement in PyMC — Write the model. Prefer PyMC 5+ syntax. Use the latest version possible.
10	4. Run prior predictive checks — `pm.sample_prior_predictive()`. Verify priors produce plausible data ranges before fitting
11	5. Inference — `pm.sample(nuts_sampler="nutpie")`. Always use nutpie for speed (the nutpie python package provides cutting-edge sampling). Don't hardcode the number of chains — let the sampler pick the best default for the platform.
12	6. Diagnose convergence — Use `arviz_stats.diagnose(idata)` as the first check (requires arviz-stats >= 1.0.0). It covers R-hat, ESS, divergences, tree depth, and E-BFMI in one call. See [references/diagnostics.md](references/diagnostics.md)
13	7. Criticize the model — See [references/model-criticism.md](references/model-criticism.md)
14	8. Check prior sensitivity — Run `psense_summary(idata)` to verify conclusions are robust to prior choices. Visualize with `plot_psense_dist(idata)` from `arviz_plots`. Requires `log_likelihood` and `log_prior` in the InferenceData — compute them after sampling if needed. See [references/sensitivity.md](references/sensitivity.md)
15	9. Compare models (if applicable) — See [references/model-comparison.md](references/model-comparison.md)
16	10. Report results — Generate `<slug>/report.md` using the canonical template in [references/reporting.md](references/reporting.md). Run `scripts/check_diagnostics.py` to turn raw diagnostics into qualitative ratings + an ordered next-steps list, and use that output to fill the Assessment lines and Suggested Next Steps section. When the user mentions a non-technical audience or is new to Bayesian stats, additionally adapt the prose to plain language and include a glossary — but keep the canonical report structure as the audit trail.
17
18	## Installation
19
20	Prefer conda-forge / mamba-forge to install PyMC and its dependencies — pip can cause issues with
21	compiled backends (nutpie, JAX). Example:
22
23	```bash
24	mamba install -c conda-forge pymc nutpie arviz arviz-stats preliz
25	```
26
27	See Stack compatibility below for the PyMC 5.x vs 6.x / ArviZ 0.23 vs 1.x notes.
28
29	## Stack compatibility (PyMC 5.x and 6.x)
30
31	This skill teaches the latest PyMC 6 / ArviZ 1.x idioms and stays runnable on
32	PyMC 5.x during the transition (regulated/corporate environments can't always
33	upgrade freely). The scripts are verified on both stacks.
34
35	Most code is identical across versions. Where an API genuinely diverges, prefer the
36	form that runs on both:
37
38	\| Task \| Modern (PyMC 6 / ArviZ 1.x) \| Legacy (PyMC 5 / ArviZ 0.23) \|
39	\|---\|---\|---\|
40	\| Posterior-predictive plot \| `arviz_plots.plot_ppc_dist(idata)` — runs on both \| `az.plot_ppc(idata)` (removed in ArviZ 1.x) \|
41	\| Trace / rank plot \| `az.plot_trace(idata, var_names=[...])`; rank view `az.plot_rank(idata, var_names=[...])` — both run on both, but pass `var_names`: ArviZ 1.x errors when the auto-selected set exceeds its subplot cap (e.g. a vector `Deterministic` like `mu` with an `obs` dim) \| `az.plot_trace(idata, kind="rank_vlines")` (the `kind=` arg is 0.23-only) \|
42	\| Prior-predictive draws \| `pm.sample_prior_predictive(draws=500)` — runs on both \| `samples=500` (removed in PyMC 6) \|
43	\| Summary interval \| `az.summary(idata, ci_prob=0.94, ci_kind="hdi")` (ArviZ 1.x only) \| `az.summary(idata, hdi_prob=0.94)` (ArviZ 0.23 only) \|
44	\| Sampler output type \| `DataTree` \| `InferenceData` \|
45
46	`arviz_plots` (the ArviZ 1.x plotting package, imported as `azp`) installs on both
47	stacks, so leading with `azp.*` plots is the most portable choice. Ba

People who liked this also viewed

6 picks

Type