DiD for Panel Data

Coverage

The simulations are based on the the make_did_SZ2020-DGP with \(1000\) observations. Learners are only set to boosting, due to time constraints (and the nonlinearity of some of the DGPs).

DoubleML Version                       0.12.dev0
Script                   did_pa_atte_coverage.py
Date                         2025-12-04 21:12:43
Total Runtime (seconds)             14567.579748
Python Version                            3.12.3

Observational Score

ⓘLearner g Learner m DGP In-sample-norm. Bias CI Length Coverage
LGBM LGBM 1 False 3.379 15.003 0.953
LGBM LGBM 2 False 3.622 17.536 0.966
LGBM LGBM 3 False 3.414 17.144 0.977
LGBM LGBM 4 False 5.765 21.603 0.932
LGBM LGBM 5 False 1.901 9.180 0.960
LGBM LGBM 6 False 1.799 9.020 0.971
LGBM LGBM 1 True 1.030 4.966 0.967
LGBM LGBM 2 True 1.224 5.982 0.963
LGBM LGBM 3 True 1.133 5.844 0.969
LGBM LGBM 4 True 1.370 7.100 0.971
LGBM LGBM 5 True 0.828 4.110 0.965
LGBM LGBM 6 True 0.800 4.053 0.970
Coverage for 95%-Confidence Interval over 1000 Repetitions
ⓘLearner g Learner m DGP In-sample-norm. Bias CI Length Coverage
LGBM LGBM 1 False 3.379 12.591 0.893
LGBM LGBM 2 False 3.622 14.717 0.914
LGBM LGBM 3 False 3.414 14.388 0.933
LGBM LGBM 4 False 5.765 18.130 0.843
LGBM LGBM 5 False 1.901 7.704 0.917
LGBM LGBM 6 False 1.799 7.570 0.922
LGBM LGBM 1 True 1.030 4.168 0.906
LGBM LGBM 2 True 1.224 5.020 0.917
LGBM LGBM 3 True 1.133 4.905 0.921
LGBM LGBM 4 True 1.370 5.959 0.925
LGBM LGBM 5 True 0.828 3.449 0.920
LGBM LGBM 6 True 0.800 3.402 0.925
Coverage for 90%-Confidence Interval over 1000 Repetitions

Experimental Score

Remark that the only two valid DGPs are DGP \(5\) and DGP \(6\). All other DGPs are invalid due to non-experimental treatment assignment.

ⓘLearner g Learner m DGP In-sample-norm. Bias CI Length Coverage
LGBM LGBM 1 False 2.170 2.577 0.094
LGBM LGBM 2 False 1.321 2.537 0.475
LGBM LGBM 3 False 1.001 2.240 0.583
LGBM LGBM 4 False 1.917 2.244 0.122
LGBM LGBM 5 False 0.521 2.460 0.951
LGBM LGBM 6 False 0.434 2.155 0.948
LGBM LGBM 1 True 2.171 2.573 0.099
LGBM LGBM 2 True 1.320 2.538 0.470
LGBM LGBM 3 True 1.000 2.239 0.598
LGBM LGBM 4 True 1.924 2.247 0.127
LGBM LGBM 5 True 0.520 2.459 0.947
LGBM LGBM 6 True 0.437 2.156 0.944
Coverage for 95%-Confidence Interval over 1000 Repetitions
ⓘLearner g Learner m DGP In-sample-norm. Bias CI Length Coverage
LGBM LGBM 1 False 2.170 2.162 0.051
LGBM LGBM 2 False 1.321 2.129 0.378
LGBM LGBM 3 False 1.001 1.880 0.470
LGBM LGBM 4 False 1.917 1.883 0.078
LGBM LGBM 5 False 0.521 2.064 0.889
LGBM LGBM 6 False 0.434 1.809 0.908
LGBM LGBM 1 True 2.171 2.160 0.049
LGBM LGBM 2 True 1.320 2.130 0.373
LGBM LGBM 3 True 1.000 1.879 0.477
LGBM LGBM 4 True 1.924 1.886 0.084
LGBM LGBM 5 True 0.520 2.064 0.891
LGBM LGBM 6 True 0.437 1.809 0.897
Coverage for 90%-Confidence Interval over 1000 Repetitions