PLIV Models

Coverage

The simulations are based on the the make_pliv_CHS2015-DGP with \(500\) observations. Due to the linearity of the DGP, Lasso is a nearly optimal choice for the nuisance estimation.

DoubleML Version                                0.12.dev0
Script                         PLIVLATECoverageSimulation
Date                                     2025-12-04 22:43
Total Runtime (minutes)                        334.088441
Python Version                                     3.12.3
Config File              scripts/plm/pliv_late_config.yml

Partialling out

ⓘLearner l Learner m Learner r Bias CI Length Coverage
LassoCV LassoCV LassoCV 0.072 0.353 0.934
LassoCV LassoCV RF Regr. 0.073 0.361 0.940
LassoCV RF Regr. LassoCV 0.075 0.372 0.945
LassoCV RF Regr. RF Regr. 0.074 0.355 0.956
RF Regr. LassoCV LassoCV 0.080 0.388 0.952
RF Regr. LassoCV RF Regr. 0.077 0.374 0.949
RF Regr. RF Regr. LassoCV 0.111 0.412 0.855
RF Regr. RF Regr. RF Regr. 0.078 0.355 0.915
Coverage for 95%-Confidence Interval over 564 Repetitions
ⓘLearner l Learner m Learner r Bias CI Length Coverage
LassoCV LassoCV LassoCV 0.072 0.296 0.885
LassoCV LassoCV RF Regr. 0.073 0.303 0.894
LassoCV RF Regr. LassoCV 0.075 0.312 0.894
LassoCV RF Regr. RF Regr. 0.074 0.298 0.906
RF Regr. LassoCV LassoCV 0.080 0.326 0.888
RF Regr. LassoCV RF Regr. 0.077 0.314 0.895
RF Regr. RF Regr. LassoCV 0.111 0.346 0.773
RF Regr. RF Regr. RF Regr. 0.078 0.298 0.848
Coverage for 90%-Confidence Interval over 564 Repetitions

IV-type

For the IV-type score, the learners ml_l and ml_g are both set to the same type of learner (here Learner g).

ⓘLearner g Learner m Learner r Bias CI Length Coverage
LassoCV LassoCV LassoCV 0.071 0.275 0.867
LassoCV LassoCV RF Regr. 0.071 0.275 0.851
LassoCV RF Regr. LassoCV 0.075 0.312 0.894
LassoCV RF Regr. RF Regr. 0.074 0.314 0.910
RF Regr. LassoCV LassoCV 0.075 0.287 0.844
RF Regr. LassoCV RF Regr. 0.076 0.286 0.856
RF Regr. RF Regr. LassoCV 0.080 0.326 0.869
RF Regr. RF Regr. RF Regr. 0.080 0.323 0.865
Coverage for 95%-Confidence Interval over 564 Repetitions
ⓘLearner g Learner m Learner r Bias CI Length Coverage
LassoCV LassoCV LassoCV 0.071 0.231 0.800
LassoCV LassoCV RF Regr. 0.071 0.230 0.789
LassoCV RF Regr. LassoCV 0.075 0.262 0.833
LassoCV RF Regr. RF Regr. 0.074 0.263 0.839
RF Regr. LassoCV LassoCV 0.075 0.241 0.787
RF Regr. LassoCV RF Regr. 0.076 0.240 0.773
RF Regr. RF Regr. LassoCV 0.080 0.273 0.791
RF Regr. RF Regr. RF Regr. 0.080 0.271 0.803
Coverage for 90%-Confidence Interval over 564 Repetitions