SSM under Nonignorable Nonresponse

Coverage

The simulations are based on the make_ssm_data-DGP with \(500\) observations. The simulation considers data with nonignorable nonresponse.

DoubleML Version                                    0.12.dev0
Script                   SSMNonIgnorableATECoverageSimulation
Date                                         2025-12-04 19:46
Total Runtime (minutes)                            156.495179
Python Version                                         3.12.3
Config File              scripts/ssm/ssm_nonig_ate_config.yml
ⓘLearner g Learner m Learner pi Bias CI Length Coverage
LGBM Regr. LGBM Clas. LGBM Clas. 0.385 1.809 0.953
LGBM Regr. LGBM Clas. Logistic 0.526 2.416 0.967
LGBM Regr. Logistic LGBM Clas. 0.314 1.334 0.897
LassoCV LGBM Clas. LGBM Clas. 0.382 1.757 0.947
LassoCV Logistic Logistic 0.479 2.077 0.920
LassoCV RF Clas. RF Clas. 0.204 0.785 0.851
RF Regr. Logistic RF Clas. 0.259 0.885 0.823
RF Regr. RF Clas. Logistic 0.398 1.780 0.953
RF Regr. RF Clas. RF Clas. 0.211 0.774 0.829
Coverage for 95%-Confidence Interval over 1000 Repetitions
ⓘLearner g Learner m Learner pi Bias CI Length Coverage
LGBM Regr. LGBM Clas. LGBM Clas. 0.385 1.518 0.894
LGBM Regr. LGBM Clas. Logistic 0.526 2.027 0.927
LGBM Regr. Logistic LGBM Clas. 0.314 1.120 0.817
LassoCV LGBM Clas. LGBM Clas. 0.382 1.474 0.875
LassoCV Logistic Logistic 0.479 1.743 0.867
LassoCV RF Clas. RF Clas. 0.204 0.659 0.771
RF Regr. Logistic RF Clas. 0.259 0.743 0.727
RF Regr. RF Clas. Logistic 0.398 1.493 0.905
RF Regr. RF Clas. RF Clas. 0.211 0.649 0.757
Coverage for 90%-Confidence Interval over 1000 Repetitions