One coefficient. Two decompositions.
Follow the note step by step, or explore any design.
What to look for
Treatment effects
You set the true effects; the design fixes the weights.
Treatment has no variation after removing the fixed effects. Change the treatment dates or add an untreated comparison group under More options.
TWFE β̂feBoth decompositions sum to this
ATTEffects weighted by Ng,t/N1
TWFE − ATT
Data
Observed group-period means
Select a treated cell to highlight it.
Two decompositions
Two weighted sums. The same TWFE coefficient.
Goodman-Bacon
Weights on 2 × 2 DiDs
β̂fe =
| Comparison | Weight × DiD | Contribution |
|---|
dCDH
Weights on effects in treated cells
β̂fe =
| Cell effect | Weight × effect | Contribution |
|---|
Diagnostics
Summaries like those of bacondecomp and twowayfeweights
Goodman-Bacon weight by comparison type
dCDH cell weights
These numbers show how exposed TWFE is to changing effects. Neither proves bias. What to do in practice →
Inspect a comparison or cell
More options: edit the design and export data
Edit the full design
| Group | Units | First treatment date tᵢ | Baseline | Initial effect | Effect change / period |
|---|
Effect after treatment = initial effect + change per period × time since treatment. Editing these values or the treatment date removes individual effect edits for that group.
Raw data table
| Group | t | Ng,t | tᵢ | D | Y(0) | Δ | Y | ε | W | W × Δ |
|---|