How to diagnose heteroskedasticity in panel data for synthetic control models. When running causal inference on panel data with synthetic controls or difference-in-differences, you need to check if residual variance differs across geos or time periods. Step 1: Visual check - plot residuals vs predicted values (funnel shape = heteroskedasticity) Step 2: Statistical test - Breusch-Pagan test confirms heteroskedasticity exists Step 3: Decompose the structure - calculate V(exp(ν)) for geo dimension

“Bell of the Divine”

TTL 3F

Behind the Scenes: Male Models at Pringle of Scotland Fall 2011 Backstage