abax 0.1.40

A lightweight Rust library providing high-precision mathematical constants and special functions, including Bernoulli numbers, Riemann Zeta values, robust incomplete gamma functions, and probability distribution functions like normal and lognormal CDF.
Documentation 
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

use crate::binopdf;

/// Binomial cumulative distribution function (CDF).
///
/// Returns the probability of obtaining `x` or fewer successes in `n` independent
/// Bernoulli trials, each with success probability `p`.
///
/// # Mathematical Definition
/// For a binomial distribution with trials `n` and probability `p`:
/// - Lower tail (`upper = false`):
/// <math display="block">
///   <mi>F</mi><mo>(</mo><mi>x</mi><mo>;</mo><mi>n</mi><mo>,</mo><mi>p</mi><mo>)</mo>
///   <mo>=</mo>
///   <munderover><mo>∑</mo><mrow><mi>i</mi><mo>=</mo><mn>0</mn></mrow><mrow><mo>⌊</mo><mi>x</mi><mo>⌋</mo></mrow></munderover>
///   <mrow><mo>(</mo><mfrac linethickness="0"><mi>n</mi><mi>i</mi></mfrac><mo>)</mo></mrow>
///   <msup><mi>p</mi><mi>i</mi></msup>
///   <msup><mrow><mo>(</mo><mn>1</mn><mo>-</mo><mi>p</mi><mo>)</mo></mrow><mrow><mi>n</mi><mo>-</mo><mi>i</mi></mrow></msup>
/// </math>
/// - Upper tail (`upper = true`):
/// <math display="block">
///   <mn>1</mn><mo>-</mo><mi>F</mi><mo>(</mo><mi>x</mi><mo>;</mo><mi>n</mi><mo>,</mo><mi>p</mi><mo>)</mo>
///   <mo>=</mo>
///   <munderover><mo>∑</mo><mrow><mi>i</mi><mo>=</mo><mo>⌊</mo><mi>x</mi><mo>⌋</mo><mo>+</mo><mn>1</mn></mrow><mi>n</mi></munderover>
///   <mrow><mo>(</mo><mfrac linethickness="0"><mi>n</mi><mi>i</mi></mfrac><mo>)</mo></mrow>
///   <msup><mi>p</mi><mi>i</mi></msup>
///   <msup><mrow><mo>(</mo><mn>1</mn><mo>-</mo><mi>p</mi><mo>)</mo></mrow><mrow><mi>n</mi><mo>-</mo><mi>i</mi></mrow></msup>
/// </math>
///
/// # Examples
/// ```
/// use abax::binocdf;
///
/// // P(2 or fewer successes in 4 trials with p=0.5)
/// // P(0)+P(1)+P(2) = 1/16 + 4/16 + 6/16 = 11/16 = 0.6875
/// assert!((binocdf(2.0, 4.0, 0.5, false) - 0.6875).abs() < 1e-15);
///
/// // Upper tail: P(X > 2) = P(3) + P(4) = 4/16 + 1/16 = 5/16 = 0.3125
/// assert!((binocdf(2.0, 4.0, 0.5, true) - 0.3125).abs() < 1e-15);
/// ```
pub fn binocdf(x: f64, n: f64, p: f64, upper: bool) -> f64 {
    let (x, p) = if upper { (n - 1.0 - f64::floor(x), 1.0 - p) } else { (x, p) };

    let mut y = 0.0;
    if x >= n {
        y = 1.0;
    }
    if x.is_nan() {
        y = f64::NAN;
    }

    let k2 = p == 0.0 && x >= 0.0;
    if k2 {
        y = 1.0;
    }

    let k3 = p == 1.0;
    if k3 {
        y = if x >= n { 1.0 } else { 0.0 };
    }

    let k1 = n < 0.0 || p < 0.0 || p > 1.0 || n != f64::round(n);
    if k1 {
        y = f64::NAN;
    }

    let xx = f64::floor(x);
    let k = xx >= 0.0 && xx < n && !k1 && !k2 && !k3;
    if k {
        let smallp = 1.0e-4;  // to maintain accuracy, don't switch tails below here
        let bign = 1.0e5;     // to conserve memory, sum over smaller range above here
        
        // for the simplest case, sum the probabilities from 0
        let np = n * p;
        let t = (n < bign) & ((x <= np) | (p < smallp));
        if t {
            y = sumfrom0(xx, n, p);
        } else if n < bign {
            y = 1.0 - sumfrom0(n - xx - 1.0, n, 1.0 - p);
        } else {
            y = sumA2B(xx, n, p);
        }
        
    }

    return y.clamp(0.0, 1.0);
}

fn sumfrom0(xx: f64, n: f64, p: f64) -> f64 {
    let y: f64 = (0..=xx as usize).map(|i| binopdf(i as f64, n, p)).sum();
    return y;
}

#[allow(non_snake_case)]
fn sumA2B(x: f64, N: f64, p: f64) -> f64 {
    let mut y: f64 = 0.0;

    // See how far we need to look into each tail
    let mu = N * p;
    let std = f64::sqrt(N * p * (1.0 - p));
    let mut t1 = 40.0;
    let mut t2 = 10.0;
    while binopdf(f64::floor(mu - t1 * std), N, p) > eps(0.0) {
        t1 = 1.5 * t1;
    }
    while binopdf(f64::ceil(mu + t2 * std), N, p) > eps(1.0) {
        t2 = 1.5 * t2;
    }

    let a = f64::max(0.0, f64::floor(mu - t1 * std));
    let b = f64::ceil(mu + t2 * std);

    if x < a {
        y = 0.0;
    }
    if x > b {
        y = 1.0;
    }
    
    if x >= a && x <= b {
        let mut sums = (a as usize .. x as usize + 1).map(|i| binopdf(i as f64, N, p)).collect::<Vec<_>>();
        for i in 1..sums.len() {
            sums[i] += sums[i - 1];
        }
        y = sums[x as usize - a as usize];
    }

    return y;
}

fn eps(x: f64) -> f64 {
    let next_up = x.next_up();
    return next_up - x;
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn test_binocdf_ex() {
        let x = 100.0;
        let n = 123456.0;
        let p = 0.001;
        let y = binocdf(x, n, p, false);
        assert!((y - 1.695618562854513e-02).abs() < 1.0e-15);
    }

    #[test]
    fn test_binocdf_basic() {
        let tol = 1e-15;
        // n=4, p=0.5, x=2 -> 0.6875
        assert!((binocdf(2.0, 4.0, 0.5, false) - 0.6875).abs() < tol);
        // n=4, p=0.5, x=2, upper -> 0.3125
        assert!((binocdf(2.0, 4.0, 0.5, true) - 0.3125).abs() < tol);
    }

    #[test]
    fn test_binocdf_boundaries() {
        // x < 0
        assert_eq!(binocdf(-1.0, 10.0, 0.5, false), 0.0);
        assert_eq!(binocdf(-1.0, 10.0, 0.5, true), 1.0);
        // x >= n
        assert_eq!(binocdf(10.0, 10.0, 0.5, false), 1.0);
        assert_eq!(binocdf(10.0, 10.0, 0.5, true), 0.0);
        assert_eq!(binocdf(11.0, 10.0, 0.5, false), 1.0);
    }

    #[test]
    fn test_binocdf_edge_p() {
        // p = 0
        assert_eq!(binocdf(0.0, 5.0, 0.0, false), 1.0);
        assert_eq!(binocdf(-1.0, 5.0, 0.0, false), 0.0);
        assert_eq!(binocdf(5.0, 5.0, 0.0, false), 1.0);
        
        // p = 1
        assert_eq!(binocdf(4.0, 5.0, 1.0, false), 0.0);
        assert_eq!(binocdf(5.0, 5.0, 1.0, false), 1.0);
        assert_eq!(binocdf(4.0, 5.0, 1.0, true), 1.0);
    }

    #[test]
    fn test_binocdf_invalid() {
        assert!(binocdf(2.0, -1.0, 0.5, false).is_nan());
        assert!(binocdf(2.0, 4.0, -0.1, false).is_nan());
        assert!(binocdf(2.0, 4.0, 1.1, false).is_nan());
        assert!(binocdf(2.0, 4.5, 0.5, false).is_nan());
        assert!(binocdf(f64::NAN, 4.0, 0.5, false).is_nan());
    }

    #[test]
    fn test_binocdf_sum_switch() {
        // Check p < smallp path
        let val = binocdf(1.0, 100.0, 1e-5, false);
        // approx P(0) + P(1) = (1-p)^n + n*p*(1-p)^(n-1)
        let p: f64 = 1e-5;
        let n: f64 = 100.0;
        let expected = (1.0 - p).powf(n) + n * p * (1.0 - p).powf(n - 1.0);
        assert!((val - expected).abs() < 1e-14);
    }

    #[test]
    fn test_binocdf_large_n() {
        // Check sumA2B path (n >= 1.0e5)
        // Reference from SciPy: scipy.stats.binom.cdf(50000, 100000, 0.5)
        // This is inclusive of the median mass.
        let val = binocdf(50000.0, 100000.0, 0.5, false);
        assert!((val - 5.012615631070982e-01).abs() < 1e-8);
    }
}