Panel Econometrics
Purpose
Estimate panel associations from entity–time tables. No-Code / Low-Code / Pro-Code views share the page; Pro-Code is not executed.
Example specification (inventory)
Pooled OLS is documented in the estimator inventory as y = a + bx + e with SVD least squares and HC1 inference. Entity fixed effects use within demeaning then SVD least squares with cluster sandwich by entity; confidence intervals are withheld below 30 clusters.
Supported runnable methods (high level)
| ID | Notes |
|---|---|
| pooled_ols, entity_fe, time_fe, twoway_fe | Operational linear projections |
| between, first_difference, random_effects, mundlak | Operational with listed limits |
| breusch_pagan, serial_correlation | Diagnostics |
| hausman, pesaran_cd, driscoll_kraay, difference_gmm, system_moments | Experimental or limited |
| anderson_hsiao | Just-identified IV; coefficient often without SE |
| arellano_bond full, blundell_bond, … | Planned — not runnable |
Inputs
CSV with entity, time, outcome, and regressor columns (built-in sample available). Optional GMM lag controls for difference GMM.
Reproducibility
Native EXACT MATCH is supported for the built-in sample entity fixed-effects path used by the notebook recorder (slope 3.000… on the fixture). Other panel methods attached as summaries are unsupported for native exact rerun.
Common mistakes
- Calling entity FE a causal effect without a design narrative.
- Interpreting withheld intervals as precision.
- Confusing identity-weight system moments with Blundell-Bond.