DID-ACTO / FIELD GUIDE
The DiD-Acto field guide
Learn what the service does, how to use it, and how to judge the evidence it returns.
A map of the service
DiD-Acto connects a research conversation to a statistical workflow. The workspace portal manages your account, uploads and saved files. Your MCP client, such as Claude or Codex, helps you inspect data, plan an analysis, run it and interpret the report. The server executes the statistical procedures and preserves the results.
MCP means Model Context Protocol: the connection through which the assistant calls the server’s tools. You do not need an SSH tunnel for this public service.
Choose a starting point
New to the service?
Start with your invitation, a shared sample and the first-run checklist.
Get started →Ready to connect?
Set up the endpoint and sign in from Claude or Codex.
Connect your assistant →Bringing a dataset?
Check the file format, panel structure and treatment definitions.
Prepare your data →Already have a result?
Read the target, uncertainty and claim status together.
Interpret the evidence →Explore the statistical toolkit
Which estimators can I use?
Find Callaway–Sant’Anna, BJS, Sun–Abraham, ETWFE, GMM, synthetic methods and all remaining registry entries with their availability.
Browse estimators →Which tests and checks exist?
Review joint pretrends, placebos, HonestDiD sensitivity, equivalence, donor diagnostics and inference methods.
Browse tests & inference →What to expect
Expect a conversation before an estimate. The tool needs a defined causal question, usable data and explicit research choices. You will review the data and comparison, approve a plan, and receive a saved result with an explanation of the estimate and checks. Missing information is resolved during planning; longer analyses may queue.
DiD-Acto is built for intervention studies. Bring the rollout history and your knowledge of the setting; use the conversation to turn those facts into explicit analysis choices.
A useful opening prompt
I am new to DiD-Acto. Explain the workflow, help me start with a shared sample, and inspect the data before planning a fit. Ask me to confirm the research question, variable roles and assumptions. Do not invent missing facts or run an analysis until I approve the plan.