Register
Connect the dataset to your question
Use a private upload or a shared sample. Give the assistant the file identifier, not a path on your laptop.
HOW IT WORKS
Your question.
A clear path to a report.
Start with a question, make the design explicit, and keep each result tied to the choices that produced it.
Why compare changes?
Imagine two groups both start with an outcome index of 100. After a program, the participating group reaches 120. Its improvement is 20 points—but the comparison group also improved by 8 points.
The simple difference-in-differences contrast is 20 minus 8, or 12 index points. It subtracts the comparison group's change from the participating group's change.
The difficult part is justifying that comparison. Without the program, would the groups have followed comparable trends? Could another change explain the difference?
INTERACTIVE TEACHING EXAMPLE
Change the two trends.
Difference in changes+12 points
20 − 8 = 12
Use these sliders to learn the arithmetic. For a fitted estimate with uncertainty and diagnostics, read the rent-control sample report.
What happens in DiD-Acto
Inspect
Understand what the data can support
Review units, periods, outcome, treatment timing and missing information. Confirm the variable roles.
Plan
Make the choices before the fit
Specify the effect you want, comparisons, assumptions, uncertainty and analysis purpose. Review the plan.
Run
Execute the approved analysis
Confirm the exact plan. The server runs the analysis in your workspace and records a durable run identifier.
Interpret
Read the evidence in context
Explain the size of the effect, how precisely it is estimated, which comparisons support it, and what the diagnostic checks found.
Report
Keep a record you can revisit
Generate a report from saved results. Reading or rendering a completed result does not fit the model again.
Review the choices that shape the result.
First review the question, variables and assumptions. Then confirm the methods and comparisons in the saved plan. After the run, review the proposed conclusion alongside the evidence.
A change to the question, outcome, comparison or assumptions starts a new plan, keeping each report tied to the choices you reviewed.
Expect questions before results.
Your assistant may need you to explain assignment, treatment timing, possible anticipation, spillovers, or the practical effect size that matters. These are research decisions; the data cannot reliably answer all of them for you.
If planning stops, the assistant should explain the missing information or unsupported design and identify the next useful step: clarify a variable, repair the table, or reconsider the comparison.