recommend-probabilistic-model

Probabilistic model selection involves mapping data signals to a method family using a decision router. This Skill diagnoses your modeling problem and recommends two to three candidate methods with a defensible default.

Yes, you can get verifiable references for Bayesian modeling decisions. The recommendation rationale cites exact Murphy §refs and pyprobml notebooks, ensuring the suggested probabilistic methods are backed by authoritative, checkable sources.

To generate a cold-start payload for a Bayesian workflow, provide your problem description and dataset details. The Skill outputs a C4 handoff payload containing likelihood families, priors, and structure for downstream execution.

The best way to compare probabilistic modeling methods is to weigh two to three candidates per problem. This approach evaluates each method against your data signals and provides a defensible default alongside a structured recommendation memo.

Yes, you need polars to run this probabilistic model selection process. The Skill relies on the polars dependency to execute its lightweight decision router and process your dataset signals for method mapping.

You should not use a decision router for probabilistic modeling when your task falls outside the eight supported deep families. If your data signals cannot be mapped to these specific families, the recommendation memo will lack a valid defensible default.