Election analysis: calculation and record checks
Election analysis: calculation and record checks
Updated September 11, 2026.
The public Python package recomputes vote counts and comparison errors from the saved official source files. The separate reference-data R package reconstructs the 78-race sample. Both were run for this update.
The original structural checker then ran to completion against a read-only snapshot. Its result was 52 of 60 checks passed. All five analytical R verifiers passed through the existing confined runner, and the instrument check passed. Eight structural assertions remain failed:
| Area | Recorded failure | Inspection of the original files |
|---|---|---|
| Research pass B plot record | Required plot log missing | The file is absent. Four chart pairs and a journal exist, but the required separate plot log does not. |
| Research pass B category map | Field declaration and unmatched-list checks fail | It uses populations and definitions, outside the checker's accepted fields structure. It records scope and exclusions, but does not supply the required enumerated categorical fields and explicit unmatched lists. |
| Model A category map | Unmatched-list check fails | Its nested field_1_... structure and distinct_raw_values_found / unmatched_after_rule names differ from the required schema. Some substantive classification details are present. The checker does not consume them. |
| Model B category map | Field declaration and unmatched-list checks fail | It records a count of 73 raw labels, classification rules and an ambiguous-label judgment, but the required enumerated values and unmatched-list fields are absent. |
| Model B plot record | Three chart references treated as missing own figures | These files exist in the two inherited research lanes. Model B explicitly labels them as inherited figures it inspected. This appears to be a false positive in the checker's extraction of every figure mention from the plot log. |
| Research pass A disclosure | 100% treated as a statistic leaked through a disclosure |
The cited passage explains why same-party general elections were excluded. It concerns the lane's own data definition, with no prior-run finding cited there. This appears to be a false positive in the disclosure heuristic. |
No assertion was weakened to obtain a passing result. The missing records and schema mismatches remain unresolved; the two likely false positives are separately documented for a future checker revision. A green reproduction result supports the calculations those scripts test, while these remaining failures concern the record and checks around the investigation.
These record deficiencies remain recorded against the original investigation. They do not constitute a successful final acceptance. Published descriptive numbers have their own runnable calculations; broader predictive validity remains unestablished.
Sources and production
Silphium is Penner Strategy’s research system. These studies use AI-assisted analysis and code-generated calculations and charts. The linked review records describe the calculation checks and remaining limitations.