# Validity Checklist - Education Data x Analysis Matrix

Use this before choosing an analysis.

## Data structure

- What is one row?
- What is the learner, group, class, artifact, or event identifier?
- Is time represented as timestamp, turn order, phase, session, or wave?
- What records are missing, filtered, or impossible for the system to capture?

## Construct mapping

- What construct is being inferred?
- What raw evidence supports that construct?
- What alternative explanations could produce the same signal?
- Is the measure a direct observation, proxy, detector output, or human interpretation?

## Method fit

- Does the method preserve the phenomenon's structure: nesting, order, relation, text, or multimodality?
- If the claim is predictive, is there leakage-free validation?
- If the claim is causal, what design feature supports causal interpretation?
- If the claim is measurement-oriented, what reliability or validity evidence exists?

## Reporting

- State the unit of analysis.
- State the claim boundary.
- Report missingness and exclusions.
- Separate observed behavior, inferred construct, and pedagogical implication.
- Name what the data cannot show.