Evaluation and metric pillars
Understand how Ounias organizes heterogeneous analytical metrics into comparable higher-level dimensions.
Why metric pillars exist
Different analytical pipelines produce different kinds of metrics.
Ounias retains pipeline-specific metrics while also organizing selected measures into broader analytical dimensions.
This provides a common layer for comparing runs without pretending that every analytical method produces identical evidence.
Metric pillars
Metrics contribute only where they meaningfully describe the corresponding analytical property.
- Separation
- Explanatory power
- Stability
- Orderliness
- Method fit
Using pillars for comparison
Pillars provide a compact common view across otherwise different runs.
They are intended to help identify strengths, weaknesses, agreement, and trade-offs across analytical configurations.
They are not a model leaderboard
Higher aggregate values do not automatically identify the scientifically correct analysis.
The pillars summarize selected analytical evidence. Interpretation still depends on the underlying metrics, method, dataset, research question, and domain context.