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
#[cfg(test)]
mod tests {
use approx::assert_relative_eq;
use scirs2_core::ndarray::array;
use crate::tests::normality::ks_2samp;
#[test]
fn test_ks_2samp_same_distribution() {
// Two samples from the same uniform distribution
let x = array![0.1, 0.2, 0.3, 0.4, 0.5];
let y = array![0.15, 0.25, 0.35, 0.45, 0.55];
let (stat, p_value) =
ks_2samp(&x.view(), &y.view(), "two-sided").expect("Test: operation failed");
// The test statistic should be reasonable for these small samples
assert!(stat <= 0.5);
// And the p-value should be large (not rejecting the null hypothesis)
assert!(p_value >= 0.01);
}
#[test]
fn test_ks_2samp_different_distributions() {
// Two samples from clearly different distributions
let x = array![0.1, 0.2, 0.3, 0.4, 0.5];
let y = array![5.1, 5.2, 5.3, 5.4, 5.5];
let (stat, p_value) =
ks_2samp(&x.view(), &y.view(), "two-sided").expect("Test: operation failed");
// The test statistic should be 1.0 (maximum difference)
assert_relative_eq!(stat, 1.0, epsilon = 1e-10);
// With such a small sample, the p-value might not be that small,
// but there should be clear evidence these are from different distributions
assert!(p_value <= 0.2);
}
#[test]
fn test_ks_2samp_one_sided_less() {
// Use larger samples for clearer results
let x = array![1.0, 1.1, 1.2, 1.3, 1.4, 1.5, 1.6, 1.7, 1.8, 1.9];
let y = array![1.5, 1.6, 1.7, 1.8, 1.9, 2.0, 2.1, 2.2, 2.3, 2.4];
// Test if x is stochastically less than y (which is true): the null
// hypothesis of the "less" test ("CDF of x >= CDF of y", i.e. x <= y)
// holds here, so we expect a HIGH p-value (fail to reject a true null).
let (_stat_less, p_value_less) =
ks_2samp(&x.view(), &y.view(), "less").expect("Test: operation failed");
// Test the opposite direction (which should be false): the null
// hypothesis of the "greater" test ("x >= y") does NOT hold here, so
// we expect a comparatively LOWER p-value (evidence against a false null).
let (_stat_greater, p_value_greater) =
ks_2samp(&x.view(), &y.view(), "greater").expect("Test: operation failed");
// A one-sided test whose null hypothesis actually matches the data
// ("less" here) should not be rejected as readily as one whose null
// is contradicted by the data ("greater" here), so its p-value
// should be the larger of the two.
assert!(
p_value_less > p_value_greater,
"p_value_less={p_value_less} should exceed p_value_greater={p_value_greater}"
);
}
#[test]
fn test_ks_2samp_one_sided_greater() {
// Use larger samples for clearer results
let y = array![1.0, 1.1, 1.2, 1.3, 1.4, 1.5, 1.6, 1.7, 1.8, 1.9];
let x = array![1.5, 1.6, 1.7, 1.8, 1.9, 2.0, 2.1, 2.2, 2.3, 2.4];
// Test if x is stochastically greater than y (which is true): the
// null hypothesis of the "greater" test ("x >= y") holds here, so we
// expect a HIGH p-value (fail to reject a true null).
let (_stat, p_value_greater) =
ks_2samp(&x.view(), &y.view(), "greater").expect("Test: operation failed");
// Test the opposite direction (which should be false): the null
// hypothesis of the "less" test ("x <= y") does NOT hold here, so we
// expect a comparatively LOWER p-value (evidence against a false null).
let (_stat, p_value_less) =
ks_2samp(&x.view(), &y.view(), "less").expect("Test: operation failed");
// This verifies that the alternative hypothesis is working correctly:
// the one-sided test whose null actually matches the data ("greater")
// should have the larger p-value of the two.
assert!(
p_value_greater > p_value_less,
"p_value_greater={p_value_greater} should exceed p_value_less={p_value_less}"
);
}
#[test]
fn test_ks_2samp_empty_arrays() {
// Empty arrays should return errors
let empty = array![];
let nonempty = array![1.0, 2.0, 3.0];
assert!(ks_2samp(&empty.view(), &nonempty.view(), "two-sided").is_err());
assert!(ks_2samp(&nonempty.view(), &empty.view(), "two-sided").is_err());
assert!(ks_2samp(&empty.view(), &empty.view(), "two-sided").is_err());
}
#[test]
fn test_ks_2samp_invalid_alternative() {
// Invalid alternative hypothesis should return an error
let x = array![1.0, 2.0, 3.0];
let y = array![4.0, 5.0, 6.0];
assert!(ks_2samp(&x.view(), &y.view(), "invalid").is_err());
}
}