File size: 4,288 Bytes
f6dac2a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
"""Day-9 tests — bootstrap estimators."""

from __future__ import annotations

import numpy as np
import pandas as pd
import pytest

from src.bootstrap import (
    BootstrapCI,
    cluster_bootstrap_did,
    cluster_bootstrap_mean_car,
    stationary_bootstrap,
)


def _make_did_panel(
    delta: float = 0.01, n_treated: int = 8, n_controls: int = 8, n_per: int = 60
) -> pd.DataFrame:
    rng = np.random.default_rng(0)
    rows: list[dict[str, float | int | str]] = []
    treated_ids = [f"T{i}" for i in range(n_treated)]
    control_ids = [f"C{i}" for i in range(n_controls)]
    for tk in treated_ids + control_ids:
        for t in range(-n_per // 2, n_per // 2):
            ret = rng.normal(0, 0.005)
            if tk in treated_ids and t >= 0:
                ret += delta
            rows.append({"ticker": tk, "t": t, "ret": ret})
    return pd.DataFrame(rows)


def test_bootstrap_ci_has_required_fields() -> None:
    ci = BootstrapCI(
        metric="x", scope="y", point=0.5, lower=0.4, upper=0.6, n_replications=100
    )
    assert ci.metric == "x" and ci.scope == "y"
    assert 0.4 <= ci.point <= 0.6
    assert ci.lower_pct == 5.0 and ci.upper_pct == 95.0


def test_cluster_bootstrap_did_recovers_delta() -> None:
    panel = _make_did_panel(delta=0.01, n_treated=8, n_controls=8, n_per=80)
    res = cluster_bootstrap_did(
        panel,
        treated={f"T{i}" for i in range(8)},
        controls={f"C{i}" for i in range(8)},
        metric_name="did_delta",
        scope="toy",
        n_reps=200,  # small for unit test speed
        seed=42,
    )
    assert res.point == pytest.approx(0.01, abs=3e-3)
    # CI should contain the true δ at 5–95
    assert res.lower < 0.01 < res.upper


def test_cluster_bootstrap_did_ci_excludes_zero_when_signal_strong() -> None:
    panel = _make_did_panel(delta=0.02, n_treated=10, n_controls=10, n_per=100)
    res = cluster_bootstrap_did(
        panel,
        treated={f"T{i}" for i in range(10)},
        controls={f"C{i}" for i in range(10)},
        metric_name="did_delta",
        scope="toy_strong",
        n_reps=300,
        seed=0,
    )
    # δ=0.02 with σ=0.005 → strong signal, 5–95 CI should not cover 0
    assert res.lower > 0


def test_stationary_bootstrap_recovers_mean() -> None:
    rng = np.random.default_rng(123)
    x = rng.normal(loc=0.05, scale=0.01, size=300)
    res = stationary_bootstrap(
        pd.Series(x),
        fn=np.mean,
        metric_name="mean",
        scope="toy",
        n_reps=500,
        mean_block_len=5,
        seed=0,
    )
    assert res.point == pytest.approx(0.05, abs=2e-3)
    assert res.lower < 0.05 < res.upper


def test_stationary_bootstrap_rejects_too_short_series() -> None:
    with pytest.raises(ValueError, match="too short"):
        stationary_bootstrap(
            pd.Series([0.0]),
            fn=np.mean,
            metric_name="m",
            scope="s",
            n_reps=10,
            seed=0,
        )


def test_cluster_bootstrap_mean_car_point_equals_sample_mean() -> None:
    """Bootstrap point estimate is the sample mean by definition."""
    rng = np.random.default_rng(7)
    cars = pd.Series(rng.normal(loc=0.02, scale=0.05, size=300))
    res = cluster_bootstrap_mean_car(
        cars, metric_name="car", scope="t1", n_reps=500, seed=0
    )
    assert res.point == pytest.approx(float(cars.mean()), abs=1e-12)
    # CI brackets the point estimate
    assert res.lower < res.point < res.upper
    # CI width is on order of 2·SE(mean) ≈ 2·0.05/√300 ≈ 0.006 → 90% CI ~0.01
    assert 0.001 < (res.upper - res.lower) < 0.05


def test_cluster_bootstrap_mean_car_rejects_empty() -> None:
    with pytest.raises(ValueError, match="empty"):
        cluster_bootstrap_mean_car(
            pd.Series([np.nan, np.nan]),
            metric_name="m",
            scope="s",
            n_reps=10,
        )


def test_cluster_bootstrap_did_rejects_empty_sets() -> None:
    panel = _make_did_panel(delta=0.0, n_treated=4, n_controls=4, n_per=50)
    with pytest.raises(ValueError, match="empty"):
        cluster_bootstrap_did(
            panel,
            treated=set(),
            controls={f"C{i}" for i in range(4)},
            metric_name="m",
            scope="s",
            n_reps=10,
        )