Reference

As-built methodology notes that other pages in this section cite as evidence: the product’s own numeric data representation, the Exhaustive search engine’s mechanism and known limitations, how its scores are calibrated into probabilities, the catalogue decisions behind the Bayesian classifier family, and the statistical primitives and hypothesis tests shared across the product’s training and analysis tools. These are code audits and design records, not R-procedure comparisons, so they sit apart from the rest of this section.

Tool Catalogue

The pages above describe this product’s statistical and model-training methodology conceptually. The semantic tool layer in Jube.Service/Agent/AdvancedAnalyticsTools – 817 individual [AdvancedAnalyticsTool("...")]-attributed methods an LLM agent calls directly by that literal name – is a separate, much larger surface that none of the pages above name tool-by-tool, and which docs/AgentHierarchy/tool-inventory.md does not cover at all (that inventory documents a different, CRUD-style tool surface). The Tool Catalogue pages in this section name every one of those 817 tools individually: what each computes, its parameters, its return shape, and real edge-case behaviour, cross-referencing a conceptual page above instead of repeating it wherever one already covers the same tool.

In this section

• Numeric Data Representation
• Shared Statistical Primitives
• Model Catalogue Decisions
• Hypothesis Tests Catalogue
• Exhaustive Methodology
• Exhaustive Calibration
• Data Preparation and Resampling Tool Catalogue
• Descriptive Statistics Tool Catalogue
• Distributions and Histograms Tool Catalogue
• Hypothesis Tests Tool Catalogue
• Relationships and Regression Tool Catalogue
• Categorical and Survival Tool Catalogue
• Multivariate and Clustering Tool Catalogue
• Time Series and Signal Tool Catalogue
• Numerics and Optimisation Tool Catalogue
• Model Evaluation Tool Catalogue
• Training Run Lifecycle Variants