Introduction to DAGassist1 months ago
Introduction | Step 0: Load DAGassist | Step 1: Declare an Estimand | Step 2: Draw a DAG | ----------------------- | Contract type (baseline; affects edu_year, income, children) | Scores: Permanent higher for higher class + urban; Informal higher for immigrant | Numeric index for linear effects (Informal=0, Temporary=1, Permanent=2) | Preference for children (baseline; affects children only) | Interpretable desired children: ~0 to 5, centered at ~2 | Centered version for linear predictors | helper: map degree to target years of schooling | ---- Try loop: redraw endogenous noise until calibration holds ---- | Job stability (time-varying; treatment-induced intermediate Z) | - affected by edu_true (treatment) and baseline contract | - affects mediators (income, married, birth_control) and outcome (children) | baseline level at t=0 | ---- Example usage ---- | Calibration check (what we target) | Attainment realism (adult-only) | Degree-type result (adult, completed degrees) | Step 3: Classify Control Variables by Role | Step 4: Estimate Models Using DAG-Consistent Adjustment Sets | Step 5: Recover the Interpretable Estimand | Export Publication-Grade Reports | Testing DGP Uncertainty with PDAGs | Testing Missing Arrows with add_edges() | References