A Practitioner’s Decision Tree for Panel-Data Estimators
Answer a handful of questions about your design and get an estimator recommendation! Each comes with assumptions, caveats, an R package, and minimal code.
What this is
This site hosts an interactive decision tree that guides you to a estimator for panel data that suits your application. It covers difference-in-differences and synthetic control method, from their simplest applications to hybrid designs and extensions. The tree branches on the structure of the design: the number of ever-treated units, treatment timing, whether the treatment is absorbing, whether it is binary or continuous. Based on your inputs, you land on a leaf with the primary recommended method, alternatives, identifying assumptions, caveats, the relevant R package, and a minimal code snippet.
The app was built as a heuristic for students working on causal inference with panel data, in response to the dizzying pace of the recent literature on difference-in-differences-style estimators and their extensions. It is a work in progress — if you find a mistake, a missing method, or have a suggestion, please open an issue.
The decision tree
The tree runs entirely in your browser via WebAssembly. The first load takes 10–20 seconds while R bootstraps; refreshing reuses the cache.
If the embedded version does not load, open the app in a new tab.
How to use it
- Answer one question at a time. Each question narrows the menu of estimators based on the structure of your data. You do not need to know the methods ahead of time.
- Land on a leaf. Each leaf returns the primary method, alternatives, identifying assumptions, caveats, an R package, and minimal example code.
- Use the breadcrumb at the top to backtrack one step, or the Start over button to reset.
- Compare a few branches. If your data structure is borderline (e.g. almost absorbing, almost common timing), walk both branches and compare the recommendations. Divergence is informative.
Citing this tool
If you use the decision tree in teaching or applied work, please cite it as:
Valli, R. (2026). A Practitioner’s Decision Tree for Panel-Data Causal Inference [online application]. https://robertovalli.github.io/practitioner_guide_did/
Acknowledgements
The tree was developed alongside the EUI workshop Advanced Counterfactual Methods for Causal Inference with Panel Data (Spring 2026). It draws on the recent econometrics literature, in particular the synthesis by Roth, Sant’Anna, Bilinski & Poe (2023), and on the implementation guidance shipped with the R packages it recommends.
License
The app and its content are released under the license in the repository.