Causal Mediation Decomposition on HMDA NY 2022 (n=30,000 stratified random sample)
7.9Estimate (pp)AIPW estimator
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| AIPW estimatorEstimand=Total effect (TE), K-fold cross-fitting=5, Nuisance models=logistic regression, Sample size (n)=30,0002026.03 | 7.9 | — | 100 | — | 0.001 | |
| AIPW estimatorEstimand=Interventional indirect effect (IIE), K-fold cross-fitting=5, Nuisance models=logistic regression, Sample size (n)=30,0002026.03 | 6.1 | 4.1 | 76.6 | — | 0.001 | |
| AIPW estimatorEstimand=Path-specific indirect effect via DTI, Estimation approach=product-of-coefficients, K-fold cross-fitting=5, Nuisance models=logistic regression, Sample size (n)=30,0002026.03 | 2.4 | — | 30.5 | — | 0.001 | |
| AIPW estimatorEstimand=Interventional direct effect (IDE), K-fold cross-fitting=5, Nuisance models=logistic regression, Sample size (n)=30,0002026.03 | 1.9 | 0.1 | 23.4 | 1.68 | 0.05 | |
| AIPW estimatorEstimand=Path-specific indirect effect via credit score quintile, Estimation approach=product-of-coefficients, K-fold cross-fitting=5, Nuisance models=logistic regression, Sample size (n)=30,0002026.03 | 1.6 | — | 20.2 | — | 0.001 | |
| AIPW estimatorEstimand=Path-specific indirect effect via income quintile, Estimation approach=product-of-coefficients, K-fold cross-fitting=5, Nuisance models=logistic regression, Sample size (n)=30,0002026.03 | 1.4 | — | 17.7 | — | 0.001 | |
| AIPW estimatorEstimand=Path-specific indirect effect via LTV, Estimation approach=product-of-coefficients, K-fold cross-fitting=5, Nuisance models=logistic regression, Sample size (n)=30,0002026.03 | 0.7 | — | 8.3 | — | 0.001 |