postforecast

Combine an existing forecast with consensus, and explain the result

postforecast takes forecasts produced elsewhere and combines them with analyst consensus. It returns an estimate and its explanation, or a refusal with a reason. Data acquisition, model training and evaluation stay in the calling application.

Start here

1. Install: Add the package or set up the notebooks.
2. Calculate: Run a complete Bayesian example.
3. Choose a method: Compare Bayesian combination and anchoring.
4. Complete a task: Publish, inspect refusals or retain an audit record.

Explore further

• Tutorials: Foundations, worked arithmetic and real-data evaluation.
• Reference: Input and output contracts, settings and API details.
• Explanation: Assumptions, provenance and design history.

What the reading date guarantees

• Availability: Preparation reads observations strictly before an explicit UTC cutoff.
• History: Bayesian statistics use earlier periods available at that cutoff.
• Provenance: Timestamps alone do not prove that an upstream model existed then.
• Calibration: Fitted settings need their own training cutoff. Research presets currently have an unknown cutoff; see preset provenance.
• Audit: Retain the input snapshot and complete settings alongside each result.