# ENA analysis decision record

Use this worksheet before creating a plot. Keep a completed copy with the analysis outputs.

## Evidence and question

- Dataset name and version:
- Dataset hash or immutable source reference:
- Relational research question:
- Why modeled co-occurrence is meaningful for this question:
- Raw-evidence access and privacy constraints:

## Segmentation and coding

- One row represents:
- Ordering field or fields:
- Code columns:
- Codebook version:
- Coding process and review evidence:
- Missing-data and exclusion rules:

## Model specification

- Unit field or fields:
- One network represents:
- Conversation field or fields (must uniquely identify a conversation across the full dataset):
- Conversation-key uniqueness check:
- Conversation-boundary rationale:
- Accumulation window and size:
- Window rationale:
- Weighting:
- Normalization:
- Rotation or projection:
- Comparison groups:
- Software and version:

## Interpretation plan

- Primary visual or numerical comparison:
- Source rows to inspect:
- Planned sensitivity checks:
- Maximum warranted claim level:
- Causal claims permitted by the design: yes / no
- Known limitations:

## After analysis

- Exported artifacts and hashes:
- Deviations from the original record:
- Patterns that changed under sensitivity checks:
- Final claim, evidence, and limitation: