Skip to content
Navigation Menu
{{ message }}
forked from igerber/diff-diff
-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathtest_imputation.py
More file actions
2168 lines (1855 loc) · 74.9 KB
/
Copy pathtest_imputation.py
File metadata and controls
2168 lines (1855 loc) · 74.9 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
"""
Tests for Borusyak-Jaravel-Spiess (2024) imputation DiD estimator.
"""
import warnings
import numpy as np
import pandas as pd
import pytest
from diff_diff.imputation import (
ImputationBootstrapResults,
ImputationDiD,
ImputationDiDResults,
imputation_did,
)
# =============================================================================
# Shared test data generation
# =============================================================================
def generate_test_data(
n_units: int = 100,
n_periods: int = 10,
treatment_effect: float = 2.0,
never_treated_frac: float = 0.3,
dynamic_effects: bool = True,
seed: int = 42,
) -> pd.DataFrame:
"""Generate synthetic staggered adoption data for testing."""
rng = np.random.default_rng(seed)
units = np.repeat(np.arange(n_units), n_periods)
times = np.tile(np.arange(n_periods), n_units)
n_never = int(n_units * never_treated_frac)
n_treated = n_units - n_never
cohort_periods = np.array([3, 5, 7])
first_treat = np.zeros(n_units, dtype=int)
if n_treated > 0:
cohort_assignments = rng.choice(len(cohort_periods), size=n_treated)
first_treat[n_never:] = cohort_periods[cohort_assignments]
first_treat_expanded = np.repeat(first_treat, n_periods)
unit_fe = rng.standard_normal(n_units) * 2.0
time_fe = np.linspace(0, 1, n_periods)
unit_fe_expanded = np.repeat(unit_fe, n_periods)
time_fe_expanded = np.tile(time_fe, n_units)
post = (times >= first_treat_expanded) & (first_treat_expanded > 0)
relative_time = times - first_treat_expanded
if dynamic_effects:
dynamic_mult = 1 + 0.1 * np.maximum(relative_time, 0)
else:
dynamic_mult = np.ones_like(relative_time, dtype=float)
effect = treatment_effect * dynamic_mult
outcomes = (
unit_fe_expanded + time_fe_expanded + effect * post + rng.standard_normal(len(units)) * 0.5
)
return pd.DataFrame(
{
"unit": units,
"time": times,
"outcome": outcomes,
"first_treat": first_treat_expanded,
}
)
# =============================================================================
# TestImputationDiD
# =============================================================================
class TestImputationDiD:
"""Tests for ImputationDiD estimator."""
def test_basic_fit(self):
"""Test basic model fitting."""
data = generate_test_data()
est = ImputationDiD()
results = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
)
assert est.is_fitted_
assert isinstance(results, ImputationDiDResults)
assert results.overall_att is not None
assert results.overall_se > 0
assert results.n_treated_obs > 0
assert results.n_untreated_obs > 0
assert results.n_treated_units > 0
assert results.n_control_units > 0
assert len(results.groups) == 3
assert len(results.time_periods) == 10
def test_positive_treatment_effect(self):
"""Test recovery of positive treatment effect."""
data = generate_test_data(treatment_effect=3.0, seed=123)
est = ImputationDiD()
results = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
)
assert results.overall_att > 0
# Effect should be close to 3.0 (dynamic effects add some)
assert abs(results.overall_att - 3.0) < 2 * results.overall_se + 1.5
def test_zero_treatment_effect(self):
"""Test with no treatment effect."""
data = generate_test_data(treatment_effect=0.0, seed=456)
est = ImputationDiD()
results = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
)
assert abs(results.overall_att) < 3 * results.overall_se + 0.5
def test_aggregate_simple(self):
"""Test that default aggregate computes overall ATT."""
data = generate_test_data()
est = ImputationDiD()
results = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
)
assert results.overall_att is not None
assert results.overall_se > 0
assert results.event_study_effects is None
assert results.group_effects is None
def test_aggregate_event_study(self):
"""Test event study aggregation."""
data = generate_test_data()
est = ImputationDiD()
results = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
aggregate="event_study",
)
assert results.event_study_effects is not None
assert len(results.event_study_effects) > 0
assert results.group_effects is None
for h, eff in results.event_study_effects.items():
assert "effect" in eff
assert "se" in eff
assert "t_stat" in eff
assert "p_value" in eff
assert "conf_int" in eff
assert "n_obs" in eff
def test_aggregate_group(self):
"""Test group aggregation."""
data = generate_test_data()
est = ImputationDiD()
results = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
aggregate="group",
)
assert results.group_effects is not None
assert len(results.group_effects) == 3 # 3 cohorts
assert results.event_study_effects is None
for g, eff in results.group_effects.items():
assert "effect" in eff
assert "se" in eff
assert eff["se"] > 0
def test_aggregate_all(self):
"""Test 'all' aggregation computes both event study and group."""
data = generate_test_data()
est = ImputationDiD()
results = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
aggregate="all",
)
assert results.event_study_effects is not None
assert results.group_effects is not None
def test_covariates(self):
"""Test estimation with covariates."""
data = generate_test_data()
rng = np.random.default_rng(99)
data["x1"] = rng.standard_normal(len(data))
data["x2"] = rng.standard_normal(len(data))
est = ImputationDiD()
results = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
covariates=["x1", "x2"],
)
assert results.overall_att is not None
assert results.overall_se > 0
def test_anticipation(self):
"""Test anticipation parameter."""
data = generate_test_data()
est0 = ImputationDiD(anticipation=0)
results0 = est0.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
aggregate="event_study",
)
est1 = ImputationDiD(anticipation=1)
results1 = est1.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
aggregate="event_study",
)
# With anticipation=1, more obs are treated, fewer untreated
assert results1.n_treated_obs > results0.n_treated_obs
# Reference period changes
ref0 = [h for h, e in results0.event_study_effects.items() if e.get("n_obs", 1) == 0]
ref1 = [h for h, e in results1.event_study_effects.items() if e.get("n_obs", 1) == 0]
assert -1 in ref0
assert -2 in ref1
def test_balance_e(self):
"""Test balance_e restricts event study to balanced cohorts."""
data = generate_test_data()
est = ImputationDiD()
results_unbal = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
aggregate="event_study",
)
results_bal = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
aggregate="event_study",
balance_e=2,
)
# Balanced should have same or fewer horizons
assert len(results_bal.event_study_effects) <= len(results_unbal.event_study_effects) + 5
def test_horizon_max(self):
"""Test horizon_max caps event study horizons."""
data = generate_test_data()
est = ImputationDiD(horizon_max=3)
results = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
aggregate="event_study",
)
for h in results.event_study_effects:
if results.event_study_effects[h].get("n_obs", 0) > 0:
assert abs(h) <= 3
def test_summary(self):
"""Test summary output."""
data = generate_test_data()
est = ImputationDiD()
results = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
aggregate="all",
)
summary = results.summary()
assert "Imputation DiD" in summary
assert "ATT" in summary
assert "Event Study" in summary
assert "Group" in summary
def test_to_dataframe_observation(self):
"""Test to_dataframe at observation level."""
data = generate_test_data()
est = ImputationDiD()
results = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
)
df = results.to_dataframe("observation")
assert "tau_hat" in df.columns
assert "weight" in df.columns
assert len(df) == results.n_treated_obs
def test_to_dataframe_event_study(self):
"""Test to_dataframe at event study level."""
data = generate_test_data()
est = ImputationDiD()
results = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
aggregate="event_study",
)
df = results.to_dataframe("event_study")
assert "relative_period" in df.columns
assert "effect" in df.columns
assert "se" in df.columns
def test_to_dataframe_group(self):
"""Test to_dataframe at group level."""
data = generate_test_data()
est = ImputationDiD()
results = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
aggregate="group",
)
df = results.to_dataframe("group")
assert "group" in df.columns
assert len(df) == 3
def test_to_dataframe_errors(self):
"""Test to_dataframe raises on invalid level."""
data = generate_test_data()
est = ImputationDiD()
results = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
)
with pytest.raises(ValueError, match="Unknown level"):
results.to_dataframe("invalid")
with pytest.raises(ValueError, match="Event study effects not computed"):
results.to_dataframe("event_study")
def test_get_params(self):
"""Test get_params returns all constructor parameters."""
est = ImputationDiD(
anticipation=1,
alpha=0.10,
n_bootstrap=100,
seed=42,
horizon_max=5,
aux_partition="cohort",
)
params = est.get_params()
assert params["anticipation"] == 1
assert params["alpha"] == 0.10
assert params["n_bootstrap"] == 100
assert params["seed"] == 42
assert params["horizon_max"] == 5
assert params["aux_partition"] == "cohort"
assert params["cluster"] is None
assert params["rank_deficient_action"] == "warn"
assert params["bootstrap_weights"] == "rademacher"
def test_bootstrap_weights_in_get_set_params(self):
"""bootstrap_weights should appear in get_params and be settable."""
est = ImputationDiD(bootstrap_weights="mammen")
assert est.get_params()["bootstrap_weights"] == "mammen"
est.set_params(bootstrap_weights="webb")
assert est.bootstrap_weights == "webb"
def test_bootstrap_weights_invalid_raises(self):
"""Invalid bootstrap_weights value should raise ValueError."""
with pytest.raises(ValueError, match="bootstrap_weights"):
ImputationDiD(bootstrap_weights="invalid")
def test_set_params(self):
"""Test set_params modifies attributes."""
est = ImputationDiD()
est.set_params(alpha=0.10, anticipation=2)
assert est.alpha == 0.10
assert est.anticipation == 2
def test_set_params_unknown(self):
"""Test set_params raises on unknown parameter."""
est = ImputationDiD()
with pytest.raises(ValueError, match="Unknown parameter"):
est.set_params(nonexistent=True)
def test_missing_columns(self):
"""Test error on missing columns."""
data = generate_test_data()
est = ImputationDiD()
with pytest.raises(ValueError, match="Missing columns"):
est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="nonexistent",
)
def test_significance_properties(self):
"""Test is_significant and significance_stars properties."""
data = generate_test_data(treatment_effect=5.0)
est = ImputationDiD()
results = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
)
assert results.is_significant
assert results.significance_stars in ("***", "**", "*", ".")
def test_repr(self):
"""Test string representation."""
data = generate_test_data()
est = ImputationDiD()
results = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
)
r = repr(results)
assert "ImputationDiDResults" in r
assert "ATT=" in r
def test_convenience_function(self):
"""Test imputation_did convenience function."""
data = generate_test_data()
results = imputation_did(
data,
"outcome",
"unit",
"time",
"first_treat",
aggregate="event_study",
)
assert isinstance(results, ImputationDiDResults)
assert results.event_study_effects is not None
def test_convenience_function_kwargs(self):
"""Test imputation_did passes kwargs to constructor."""
data = generate_test_data()
results = imputation_did(
data,
"outcome",
"unit",
"time",
"first_treat",
alpha=0.10,
)
assert results.alpha == 0.10
def test_unbalanced_panel(self):
"""Test with unbalanced panel (some units missing periods)."""
data = generate_test_data(seed=99)
rng = np.random.default_rng(99)
# Drop some observations randomly
keep = rng.random(len(data)) > 0.1
data_unbal = data[keep].reset_index(drop=True)
est = ImputationDiD()
results = est.fit(
data_unbal,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
)
assert results.overall_att is not None
assert results.overall_se > 0
def test_balance_e_checks_pre_treatment_periods(self):
"""balance_e should drop cohorts missing pre-treatment observations."""
# Cohort A (first_treat=4): units 0-4, all periods 0-7
# rel_times: -4, -3, -2, -1, 0, 1, 2, 3
# Cohort B (first_treat=6): units 5-9, all periods 0-7 EXCEPT time=4
# rel_times: -6, -5, -4, -3, -1, 0, 1 (missing -2)
# Never-treated: units 10-14, all periods
#
# horizon_max=1 caps post-treatment to {0,1} so both cohorts can
# cover the required post-treatment range. Without it, the union of
# all_horizons includes h=2,3 which cohort B can't reach (max h=1).
rows = []
rng = np.random.default_rng(123)
# Cohort A: complete panel
for u in range(5):
for t in range(8):
y = u * 0.5 + t * 0.1 + (3.0 if t >= 4 else 0.0)
rows.append(
{
"unit": u,
"time": t,
"first_treat": 4,
"outcome": y + rng.normal(0, 0.01),
}
)
# Cohort B: missing time=4 (which is rel_time = 4 - 6 = -2)
for u in range(5, 10):
for t in range(8):
if t == 4:
continue # drop => missing rel_time=-2
y = u * 0.5 + t * 0.1 + (3.0 if t >= 6 else 0.0)
rows.append(
{
"unit": u,
"time": t,
"first_treat": 6,
"outcome": y + rng.normal(0, 0.01),
}
)
# Never-treated
for u in range(10, 15):
for t in range(8):
y = u * 0.5 + t * 0.1
rows.append(
{
"unit": u,
"time": t,
"first_treat": 0,
"outcome": y + rng.normal(0, 0.01),
}
)
data = pd.DataFrame(rows)
est = ImputationDiD(horizon_max=1)
# balance_e=2, horizon_max=1: required = {-2,-1,0,1}
# Cohort B missing -2 => should be dropped
results_bal2 = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
aggregate="event_study",
balance_e=2,
)
# balance_e=1, horizon_max=1: required = {-1,0,1}
# Both cohorts have -1 => both kept
results_bal1 = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
aggregate="event_study",
balance_e=1,
)
# Cohort B dropped at balance_e=2 => fewer obs at horizon 0
n_obs_bal2_h0 = results_bal2.event_study_effects[0]["n_obs"]
n_obs_bal1_h0 = results_bal1.event_study_effects[0]["n_obs"]
assert n_obs_bal2_h0 < n_obs_bal1_h0, (
f"balance_e=2 should drop cohort B (missing rel_time=-2), "
f"got n_obs={n_obs_bal2_h0} vs {n_obs_bal1_h0}"
)
# =============================================================================
# TestImputationDiDResults
# =============================================================================
class TestImputationDiDResults:
"""Tests for ImputationDiDResults."""
def test_pretrend_test(self):
"""Test pre-trend test on data with parallel trends."""
data = generate_test_data(dynamic_effects=False, seed=77, n_units=200)
est = ImputationDiD()
results = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
)
pt = results.pretrend_test()
assert "f_stat" in pt
assert "p_value" in pt
assert "n_leads" in pt
assert pt["n_leads"] > 0
# Under parallel trends, should not reject
assert pt["p_value"] > 0.01
def test_pretrend_with_violation(self):
"""Test pre-trend test detects trend violation."""
data = generate_test_data(seed=88, n_units=200)
# Add a pre-treatment trend for treated units
rng = np.random.default_rng(88)
for idx in data.index:
if data.loc[idx, "first_treat"] > 0:
t = data.loc[idx, "time"]
ft = data.loc[idx, "first_treat"]
if t < ft:
data.loc[idx, "outcome"] += 0.5 * (t - ft)
est = ImputationDiD()
results = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
)
pt = results.pretrend_test()
# With pre-trend violation, should reject (low p-value)
assert pt["p_value"] < 0.10
def test_pretrend_unbalanced_panel(self):
"""Test pretrend_test uses iterative demeaning for unbalanced panels."""
data = generate_test_data(dynamic_effects=False, seed=77, n_units=200)
# Make unbalanced by dropping ~15% of observations
rng = np.random.default_rng(77)
keep = rng.random(len(data)) > 0.15
data_unbal = data[keep].reset_index(drop=True)
est = ImputationDiD()
results = est.fit(
data_unbal,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
)
pt = results.pretrend_test()
assert pt["n_leads"] > 0
# Under parallel trends, should not reject
assert pt["p_value"] > 0.01
def test_pretrend_n_leads(self):
"""Test pre-trend test with specified number of leads."""
data = generate_test_data(n_units=200, seed=55)
est = ImputationDiD()
results = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
)
pt = results.pretrend_test(n_leads=2)
assert pt["n_leads"] == 2
# =============================================================================
# TestImputationVariance
# =============================================================================
class TestImputationVariance:
"""Tests for conservative variance estimation (Theorem 3)."""
def test_se_positive(self):
"""Test that SE is positive."""
data = generate_test_data()
est = ImputationDiD()
results = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
)
assert results.overall_se > 0
def test_se_positive_event_study(self):
"""Test that event study SEs are positive."""
data = generate_test_data()
est = ImputationDiD()
results = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
aggregate="event_study",
)
for h, eff in results.event_study_effects.items():
if eff.get("n_obs", 0) > 0 and np.isfinite(eff["se"]):
assert eff["se"] > 0
def test_aux_partition_cohort_horizon(self):
"""Test cohort_horizon partition produces valid SEs."""
data = generate_test_data()
est = ImputationDiD(aux_partition="cohort_horizon")
results = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
)
assert results.overall_se > 0
def test_aux_partition_cohort(self):
"""Test cohort partition produces valid SEs."""
data = generate_test_data()
est = ImputationDiD(aux_partition="cohort")
results = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
)
assert results.overall_se > 0
def test_aux_partition_horizon(self):
"""Test horizon partition produces valid SEs."""
data = generate_test_data()
est = ImputationDiD(aux_partition="horizon")
results = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
)
assert results.overall_se > 0
def test_coarser_partition_more_conservative(self):
"""Test that coarser partition gives more conservative (larger) SEs."""
data = generate_test_data(n_units=200, seed=42)
est_fine = ImputationDiD(aux_partition="cohort_horizon")
results_fine = est_fine.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
)
est_coarse = ImputationDiD(aux_partition="cohort")
results_coarse = est_coarse.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
)
# Coarser partition should give >= SE (approximately)
# Allow small tolerance for numerical issues
assert results_coarse.overall_se >= results_fine.overall_se * 0.95
def test_invalid_aux_partition(self):
"""Test that invalid aux_partition raises ValueError."""
with pytest.raises(ValueError, match="aux_partition"):
ImputationDiD(aux_partition="invalid")
def test_sparse_solver_matches_dense(self):
"""Test that sparse solver produces finite SEs with covariates."""
data = generate_test_data(n_units=100, n_periods=10, seed=42)
rng = np.random.default_rng(42)
data["x1"] = rng.standard_normal(len(data))
est = ImputationDiD()
results = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
covariates=["x1"],
)
assert np.isfinite(results.overall_se)
assert results.overall_se > 0
def test_sparse_solver_dense_fallback(self):
"""Test that dense fallback produces finite SE when spsolve fails."""
import unittest.mock
data = generate_test_data(n_units=80, n_periods=8, seed=42)
rng = np.random.default_rng(42)
data["x1"] = rng.standard_normal(len(data))
est = ImputationDiD()
# Monkey-patch spsolve to force fallback to dense lstsq
with unittest.mock.patch(
"diff_diff.imputation.spsolve", side_effect=RuntimeError("test failure")
):
results = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
covariates=["x1"],
)
assert np.isfinite(results.overall_se)
assert results.overall_se > 0
def test_sparse_solver_dense_fallback_emits_warning(self):
"""Silent-failure audit axis C: the sparse -> dense lstsq fallback must
emit a UserWarning so callers are informed that variance estimates come
from the degraded path."""
import unittest.mock
data = generate_test_data(n_units=80, n_periods=8, seed=42)
rng = np.random.default_rng(42)
data["x1"] = rng.standard_normal(len(data))
est = ImputationDiD()
with unittest.mock.patch(
"diff_diff.imputation.spsolve", side_effect=RuntimeError("test failure")
):
with pytest.warns(UserWarning, match="sparse solve.*falling back to dense lstsq"):
est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
covariates=["x1"],
)
# =============================================================================
# TestImputationBootstrap
# =============================================================================
class TestImputationBootstrap:
"""Tests for bootstrap inference."""
def test_basic_bootstrap(self, ci_params):
"""Test basic bootstrap inference."""
data = generate_test_data()
n_boot = ci_params.bootstrap(50)
est = ImputationDiD(n_bootstrap=n_boot, seed=42)
results = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
)
assert results.bootstrap_results is not None
assert isinstance(results.bootstrap_results, ImputationBootstrapResults)
assert results.bootstrap_results.n_bootstrap == n_boot
assert results.bootstrap_results.overall_att_se > 0
def test_bootstrap_reproducibility(self, ci_params):
"""Test that same seed gives same results."""
data = generate_test_data()
n_boot = ci_params.bootstrap(50)
est1 = ImputationDiD(n_bootstrap=n_boot, seed=42)
r1 = est1.fit(data, outcome="outcome", unit="unit", time="time", first_treat="first_treat")
est2 = ImputationDiD(n_bootstrap=n_boot, seed=42)
r2 = est2.fit(data, outcome="outcome", unit="unit", time="time", first_treat="first_treat")
assert r1.overall_se == r2.overall_se
def test_bootstrap_different_seeds(self, ci_params):
"""Test that different seeds give different results."""
data = generate_test_data()
n_boot = ci_params.bootstrap(50)
est1 = ImputationDiD(n_bootstrap=n_boot, seed=42)
r1 = est1.fit(data, outcome="outcome", unit="unit", time="time", first_treat="first_treat")
est2 = ImputationDiD(n_bootstrap=n_boot, seed=99)
r2 = est2.fit(data, outcome="outcome", unit="unit", time="time", first_treat="first_treat")
# Results should differ (at least slightly)
assert r1.overall_se != r2.overall_se
def test_bootstrap_event_study(self, ci_params):
"""Test bootstrap with event study aggregation."""
data = generate_test_data()
n_boot = ci_params.bootstrap(50)
est = ImputationDiD(n_bootstrap=n_boot, seed=42)
results = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
aggregate="event_study",
)
br = results.bootstrap_results
assert br.event_study_ses is not None
assert len(br.event_study_ses) > 0
def test_bootstrap_group(self, ci_params):
"""Test bootstrap with group aggregation."""
data = generate_test_data()
n_boot = ci_params.bootstrap(50)
est = ImputationDiD(n_bootstrap=n_boot, seed=42)
results = est.fit(
data,
outcome="outcome",
unit="unit",
time="time",
first_treat="first_treat",
aggregate="group",
)
br = results.bootstrap_results
assert br.group_ses is not None
You can’t perform that action at this time.