GATEs

Coverage

The simulations are based on the the make_heterogeneous_data-DGP with \(500\) observations. The groups are defined based on the first covariate, analogously to the GATE IRM Example, but rely on LightGBM to estimate nuisance elements (due to time constraints).

The non-uniform results (coverage, ci length and bias) refer to averaged values over all groups (point-wise confidende intervals).

DoubleML Version                            0.11.5.dev55
Script                         IRMGATECoverageSimulation
Date                                    2026-08-10 11:53
Total Runtime (minutes)                        80.167192
Python Version                                    3.12.3
Config File              scripts/irm/irm_gate_config.yml
ⓘLearner g Learner m Bias CI Length Coverage Uniform CI Length Uniform Coverage
LGBM Regr. LGBM Clas. 0.189 1.015 0.974 1.236 0.978
LGBM Regr. Logistic 0.096 0.478 0.949 0.582 0.949
Linear LGBM Clas. 0.193 1.011 0.964 1.230 0.975
Linear Logistic 0.098 0.499 0.955 0.607 0.960
Linear RF Clas. 0.100 0.527 0.965 0.642 0.969
RF Regr. Logistic 0.096 0.477 0.958 0.581 0.951
RF Regr. RF Clas. 0.100 0.503 0.955 0.613 0.955
Coverage for 95%-Confidence Interval over 1000 Repetitions
ⓘLearner g Learner m Bias CI Length Coverage Uniform CI Length Uniform Coverage
LGBM Regr. LGBM Clas. 0.189 0.852 0.932 1.235 0.976
LGBM Regr. Logistic 0.096 0.401 0.896 0.581 0.954
Linear LGBM Clas. 0.193 0.848 0.919 1.229 0.977
Linear Logistic 0.098 0.418 0.909 0.607 0.962
Linear RF Clas. 0.100 0.442 0.925 0.641 0.969
RF Regr. Logistic 0.096 0.400 0.904 0.581 0.954
RF Regr. RF Clas. 0.100 0.422 0.907 0.613 0.957
Coverage for 90%-Confidence Interval over 1000 Repetitions