Connectedness feature reference
A Shannon-entropy measure of how evenly the focal place's figure relationships are distributed across canonical action categories.
Stored model field:
action_profile_entropy
Count distinct figures in each action category, convert the counts to shares, and calculate base-two Shannon entropy. Empty or one-category profiles receive zero.
The action profile is more varied and more evenly spread across categories.
These descriptive statistics use the current labelled cohort after aliases have been collapsed to one row per MANTO place.
| Group | N | Minimum | Q1 | Median | Mean | Q3 | Maximum |
|---|---|---|---|---|---|---|---|
| All labelled places | 255 | 0.000 | 0.000 | 0.000 | 0.270 | 0.000 | 3.406 |
| Survives | 204 | 0.000 | 0.000 | 0.000 | 0.280 | 0.000 | 2.928 |
| Does not survive | 51 | 0.000 | 0.000 | 0.000 | 0.233 | 0.000 | 3.406 |
Examples are descriptive checks on the calculated feature. Their outcome labels report what Pausanias says; they do not establish that the feature caused survival or abandonment.
| Place | MANTO ID | Value | Pausanias reports |
|---|---|---|---|
| 🌍 Thebes (Boiotia) | 8253960 | 3.406 | Does not survive |
| 🌍 Aigina (city) | 10191987 | 2.928 | Survives |
| 🌍 Argos (city) | 8228853 | 2.775 | Survives |
| 🌍 Syracuse (Sicily) | 11295783 | 2.753 | Survives |
| 🌍 Opous (Locris) | 8253989 | 2.522 | Survives |
| 🌍 Rome (Italy) | 10266956 | 2.333 | Survives |
| Place | MANTO ID | Value | Pausanias reports |
|---|---|---|---|
| 🌍 Abai (Phocis) | 10274687 | 0.000 | Survives |
| 🌍 Abia (Messenia) | 10113505 | 0.000 | Survives |
| 🌍 Acacesion (Arcadia) | 10126185 | 0.000 | Survives |
| 🌍 Ace (Arcadia) | 10214517 | 0.000 | Survives |
| 🌍 Acharnai (Attica) | 10153056 | 0.000 | Survives |
| 🌍 Acriai (Laconia) | 10149418 | 0.000 | Survives |
Its current standardized coefficient is +0.272, pointing toward survives. This means that a one-standard-deviation increase changes the fitted log odds in that direction while the other included features are held fixed.
Because the connectedness features are correlated and the model's out-of-fold performance is weak, this is a conditional association—not a standalone ranking, causal effect, or historical explanation.