Tutorials

Learn by running a complete worked example

Before starting

  • Installation: The package is internal to Exabel and installed from git. Replace <commit> with the immutable revision selected for the project:

    uv add 'postforecast @ git+https://github.com/Exabel/postforecast.git@<commit>'

  • Python: Supported versions are 3.11–3.13.

  • Notebook environment: In a repository checkout, run just install and select .venv/bin/python as the notebook kernel. Keep docs/pf.py beside the tutorial notebooks; it supplies their example data and figure styling.

  • Reading online: Every notebook includes computed outputs, so the walkthroughs can also be followed without running a local environment.

Follow the learning path

  1. The panel: Build and validate the long estimate frame. Observe how invalid timestamps and ambiguous observations are rejected.
  2. Point-in-time: Read values before an explicit moment and explore how staleness and training boundaries affect the result.
  3. Anchoring: Work from forecasts and consensus through the revision, weighted surprise, refusal checks and versioned estimate.
  4. Bayesian posterior: Assemble historical growth and error statistics, combine the readings and inspect the uncertainty interval.

Continue with a real task

  • Publishing: Use the how-to guides for the high-level panel APIs.
  • Complete Bayesian workflow: The Bayesian notebook teaches BayesianPosterior. Use Bayesian prediction configuration for bayesian_kpi_predictions, including the correlation-adjusted method.
  • Mathematics: Consult the reference for equations, defaults and reference-specific behavior.
  • Rationale: Follow the explanation links for the assumptions behind the calculations.