solvr 0.2.0

Advanced computing library for real-world problem solving - optimization, differential equations, interpolation, statistics, and more
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
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

//! CPU implementation of hypothesis testing algorithms.

use crate::stats::TensorTestResult;
use crate::stats::impl_generic::{
    bartlett_impl, f_oneway_impl, friedmanchisquare_impl, kruskal_impl, levene_impl,
    normaltest_impl, pearsonr_impl, shapiro_impl, spearmanr_impl, ttest_1samp_impl, ttest_ind_impl,
    ttest_rel_impl,
};
use crate::stats::traits::{HypothesisTestingAlgorithms, LeveneCenter};
use numr::error::Result;
use numr::runtime::cpu::{CpuClient, CpuRuntime};
use numr::tensor::Tensor;

impl HypothesisTestingAlgorithms<CpuRuntime> for CpuClient {
    fn ttest_1samp(
        &self,
        x: &Tensor<CpuRuntime>,
        popmean: f64,
    ) -> Result<TensorTestResult<CpuRuntime>> {
        ttest_1samp_impl(self, x, popmean)
    }

    fn ttest_ind(
        &self,
        a: &Tensor<CpuRuntime>,
        b: &Tensor<CpuRuntime>,
    ) -> Result<TensorTestResult<CpuRuntime>> {
        ttest_ind_impl(self, a, b)
    }

    fn ttest_rel(
        &self,
        a: &Tensor<CpuRuntime>,
        b: &Tensor<CpuRuntime>,
    ) -> Result<TensorTestResult<CpuRuntime>> {
        ttest_rel_impl(self, a, b)
    }

    fn pearsonr(
        &self,
        x: &Tensor<CpuRuntime>,
        y: &Tensor<CpuRuntime>,
    ) -> Result<TensorTestResult<CpuRuntime>> {
        pearsonr_impl(self, x, y)
    }

    fn spearmanr(
        &self,
        x: &Tensor<CpuRuntime>,
        y: &Tensor<CpuRuntime>,
    ) -> Result<TensorTestResult<CpuRuntime>> {
        spearmanr_impl(self, x, y)
    }

    fn f_oneway(&self, groups: &[&Tensor<CpuRuntime>]) -> Result<TensorTestResult<CpuRuntime>> {
        f_oneway_impl(self, groups)
    }

    fn kruskal(&self, groups: &[&Tensor<CpuRuntime>]) -> Result<TensorTestResult<CpuRuntime>> {
        kruskal_impl(self, groups)
    }

    fn friedmanchisquare(
        &self,
        groups: &[&Tensor<CpuRuntime>],
    ) -> Result<TensorTestResult<CpuRuntime>> {
        friedmanchisquare_impl(self, groups)
    }

    fn shapiro(&self, x: &Tensor<CpuRuntime>) -> Result<TensorTestResult<CpuRuntime>> {
        shapiro_impl(self, x)
    }

    fn normaltest(&self, x: &Tensor<CpuRuntime>) -> Result<TensorTestResult<CpuRuntime>> {
        normaltest_impl(self, x)
    }

    fn levene(
        &self,
        groups: &[&Tensor<CpuRuntime>],
        center: LeveneCenter,
    ) -> Result<TensorTestResult<CpuRuntime>> {
        levene_impl(self, groups, center)
    }

    fn bartlett(&self, groups: &[&Tensor<CpuRuntime>]) -> Result<TensorTestResult<CpuRuntime>> {
        bartlett_impl(self, groups)
    }
}

#[cfg(test)]
mod tests {
    use super::*;
    use crate::stats::helpers::extract_scalar;
    use numr::runtime::cpu::CpuDevice;

    fn setup() -> (CpuClient, CpuDevice) {
        let device = CpuDevice::new();
        let client = CpuClient::new(device.clone());
        (client, device)
    }

    #[test]
    fn test_ttest_1samp() {
        let (client, device) = setup();
        let data = Tensor::<CpuRuntime>::from_slice(&[1.2f64, 1.5, 1.3, 1.4, 1.6], &[5], &device);

        let result = client.ttest_1samp(&data, 1.0).unwrap();
        let stat = extract_scalar(&result.statistic).unwrap();
        let pval = extract_scalar(&result.pvalue).unwrap();

        assert!(stat > 0.0);
        assert!(pval < 0.05);
    }

    #[test]
    fn test_ttest_ind() {
        let (client, device) = setup();
        let a = Tensor::<CpuRuntime>::from_slice(&[1.0f64, 2.0, 3.0, 4.0, 5.0], &[5], &device);
        let b = Tensor::<CpuRuntime>::from_slice(&[2.0f64, 3.0, 4.0, 5.0, 6.0], &[5], &device);

        let result = client.ttest_ind(&a, &b).unwrap();
        let stat = extract_scalar(&result.statistic).unwrap();
        let pval = extract_scalar(&result.pvalue).unwrap();

        assert!(stat < 0.0);
        assert!(pval > 0.0 && pval < 1.0);
    }

    #[test]
    fn test_pearsonr() {
        let (client, device) = setup();
        let x = Tensor::<CpuRuntime>::from_slice(&[1.0f64, 2.0, 3.0, 4.0, 5.0], &[5], &device);
        let y = Tensor::<CpuRuntime>::from_slice(&[2.0f64, 4.0, 6.0, 8.0, 10.0], &[5], &device);

        let result = client.pearsonr(&x, &y).unwrap();
        let r = extract_scalar(&result.statistic).unwrap();
        let pval = extract_scalar(&result.pvalue).unwrap();

        assert!((r - 1.0).abs() < 1e-10);
        assert!(pval < 0.01);
    }

    #[test]
    fn test_f_oneway() {
        let (client, device) = setup();
        let a = Tensor::<CpuRuntime>::from_slice(&[1.0f64, 2.0, 3.0, 4.0, 5.0], &[5], &device);
        let b = Tensor::<CpuRuntime>::from_slice(&[2.0f64, 3.0, 4.0, 5.0, 6.0], &[5], &device);
        let c = Tensor::<CpuRuntime>::from_slice(&[10.0f64, 11.0, 12.0, 13.0, 14.0], &[5], &device);

        let result = client.f_oneway(&[&a, &b, &c]).unwrap();
        let f_stat = extract_scalar(&result.statistic).unwrap();
        let pval = extract_scalar(&result.pvalue).unwrap();

        // Group c is very different, should have high F and low p
        assert!(f_stat > 1.0);
        assert!(pval < 0.01);
    }

    #[test]
    fn test_kruskal() {
        let (client, device) = setup();
        let a = Tensor::<CpuRuntime>::from_slice(&[1.0f64, 2.0, 3.0, 4.0, 5.0], &[5], &device);
        let b = Tensor::<CpuRuntime>::from_slice(&[10.0f64, 11.0, 12.0, 13.0, 14.0], &[5], &device);

        let result = client.kruskal(&[&a, &b]).unwrap();
        let h = extract_scalar(&result.statistic).unwrap();
        let pval = extract_scalar(&result.pvalue).unwrap();

        assert!(h > 0.0);
        assert!(pval < 0.05);
    }

    #[test]
    fn test_friedmanchisquare() {
        let (client, device) = setup();
        // 5 subjects, 3 treatments
        let t1 = Tensor::<CpuRuntime>::from_slice(&[1.0f64, 2.0, 3.0, 4.0, 5.0], &[5], &device);
        let t2 = Tensor::<CpuRuntime>::from_slice(&[2.0f64, 3.0, 4.0, 5.0, 6.0], &[5], &device);
        let t3 =
            Tensor::<CpuRuntime>::from_slice(&[10.0f64, 11.0, 12.0, 13.0, 14.0], &[5], &device);

        let result = client.friedmanchisquare(&[&t1, &t2, &t3]).unwrap();
        let chi2 = extract_scalar(&result.statistic).unwrap();
        let pval = extract_scalar(&result.pvalue).unwrap();

        assert!(chi2 > 0.0);
        assert!(pval < 0.05);
    }

    #[test]
    fn test_shapiro() {
        let (client, device) = setup();
        // Data from a normal distribution
        let data = Tensor::<CpuRuntime>::from_slice(
            &[-1.2f64, -0.5, 0.0, 0.3, 0.7, 1.0, 1.5, 2.0, -0.8, 0.2],
            &[10],
            &device,
        );

        let result = client.shapiro(&data).unwrap();
        let w = extract_scalar(&result.statistic).unwrap();
        let pval = extract_scalar(&result.pvalue).unwrap();

        // W should be close to 1 for normal data
        assert!(w > 0.8);
        // p-value should not reject normality
        assert!(pval > 0.01);
    }

    #[test]
    fn test_normaltest() {
        let (client, device) = setup();
        // Roughly normal data (n=30)
        let data: Vec<f64> = (0..30).map(|i| (i as f64 - 15.0) / 5.0).collect();
        let x = Tensor::<CpuRuntime>::from_slice(&data, &[30], &device);

        let result = client.normaltest(&x).unwrap();
        let k2 = extract_scalar(&result.statistic).unwrap();
        let pval = extract_scalar(&result.pvalue).unwrap();

        assert!(k2 >= 0.0);
        assert!((0.0..=1.0).contains(&pval));
    }

    #[test]
    fn test_levene() {
        let (client, device) = setup();
        let a = Tensor::<CpuRuntime>::from_slice(&[1.0f64, 2.0, 3.0, 4.0, 5.0], &[5], &device);
        let b = Tensor::<CpuRuntime>::from_slice(&[1.0f64, 10.0, 1.0, 10.0, 1.0], &[5], &device);

        // b has much higher variance
        let result = client.levene(&[&a, &b], LeveneCenter::Median).unwrap();
        let f_stat = extract_scalar(&result.statistic).unwrap();
        let pval = extract_scalar(&result.pvalue).unwrap();

        assert!(f_stat > 0.0);
        assert!((0.0..=1.0).contains(&pval));
    }

    #[test]
    fn test_bartlett() {
        let (client, device) = setup();
        let a = Tensor::<CpuRuntime>::from_slice(&[1.0f64, 2.0, 3.0, 4.0, 5.0], &[5], &device);
        let b =
            Tensor::<CpuRuntime>::from_slice(&[1.0f64, 10.0, 1.0, 10.0, 1.0, 10.0], &[6], &device);

        let result = client.bartlett(&[&a, &b]).unwrap();
        let t_stat = extract_scalar(&result.statistic).unwrap();
        let pval = extract_scalar(&result.pvalue).unwrap();

        assert!(t_stat > 0.0);
        assert!((0.0..=1.0).contains(&pval));
    }
}