Tutorials

Learn the calculation, then apply it to real data

Start with Installation and Usage. Saved notebook outputs can be read without a local environment.

Foundations

1. Terminology: Forecasts, subjects, growth and reading dates.
2. The input panel: Build and validate observations.
3. Point-in-time: Reading cutoffs and historical evaluation.

Worked calculations

• Bayesian posterior: Precisions, weights, intervals and refusals on a small in-memory example.
• Anchoring: Frozen coefficients, contributions and eligibility on a small in-memory example.

Real-data applications

1. Real model walkthrough: Inspect one company’s inputs and explain both estimates. Assumes familiarity with a worked calculation.
2. Evaluate forecast improvement: Replay historical reading dates and compare errors. Assumes familiarity with the walkthrough.

Both real-data notebooks require a local ADC Parquet export; see notebook setup. All other notebooks build their inputs in memory. Export-vintage limitations are stated beside the examples, so an illustrative replay is not mistaken for a live out-of-sample trial.