DID-ACTO / FIELD GUIDE
Understand your results
Read the estimate, understand the checks, and explain what the study found.
Read in this order
- Target: which effect, population and periods?
- Scale: outcome units, percentage points, levels or another declared transformation?
- Estimate and uncertainty: point estimate, interval or band?
- Support: which units and comparisons contribute?
- Diagnostics and sensitivity: what challenges the interpretation?
- Claim status: what wording is actually supported?
Read the size and precision together
Read the confidence interval as a range of effects compatible with the estimate and its statistical model at the stated confidence level. A narrow interval signals greater precision; assess the comparison design separately.
If the interval spans zero, the data may still be consistent with effects large enough to matter. Compare both endpoints with a practical threshold chosen before the analysis. In the rent-control sample, the dollar scale makes the estimated change easy to interpret.
Read checks in context
Start with the pre-treatment pattern: do the groups move similarly, and how much information is available to detect a difference? Read the joint test alongside the plot. Then use supported sensitivity analysis to see how the interval changes under specified departures from parallel trends.
If methods disagree, compare their targets, included observations, weights and assumptions. Ask which difference explains the gap before deciding what to emphasize.
Know the difference between statuses
| Status or situation | What it means for you |
|---|---|
| Queued / running | The analysis is not complete. Check the same run identifier. |
| Completed | Computation finished. Review evidence and claim eligibility before interpreting it causally. |
| Descriptive / unsupported claim | Describe the observed comparison and explain which design or evidence requirement is still unresolved. |
| Inconclusive evidence | Uncertainty or design limitations leave the substantive question unresolved. |
| Quarantined / unavailable | Integrity or compatibility checks prevent normal access. Ask an administrator to investigate; do not bypass the verifier. |
Exact status fields depend on the workflow. The status returned with your saved result is authoritative.
Download and revisit
Use Files and reports → Refresh library in the portal to locate saved analyses. Available report downloads are generated through the verified result path. MCP can also retrieve the result or render the report using the existing run identifier.
Rendering a completed report uses stored results and does not refit. If you want to change the specification, start a new plan and keep its new run distinct from the previous one.
Turn the report into a useful summary
Explain this saved report in five parts: the study question; the effect and its units; the uncertainty; what the diagnostic checks found; and the most useful next step. Keep the report’s evidence status attached to the conclusion.