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

//! Run required statistical tests over data
use std::cmp::Eq;
use std::collections::HashMap;
use std::hash::Hash;

/// Simple mean
pub fn avg(l: &mut impl std::iter::Iterator<Item = f32>) -> f32 {
    let mut sum = 0.0;
    let mut len = 0;
    for n in l {
        sum += n;
        len += 1;
    }

    sum / len as f32
}

/// Sample standard deviation (sum of suared differences / dof)
pub fn stdev(
    l: &mut (impl std::iter::Iterator<Item = f32> + Clone)
) -> Option<f32> {
    let x_bar = avg(&mut l.clone());
    let len = l.clone().collect::<Vec<f32>>().len();

    match len {
        0..=1 => None,
        len => Some(
            l.map(|x| (x - x_bar).powf(2.0)).sum::<f32>() / len as f32
        ),
    }
}

/// Given a hashmap of ((T, T), Vec<f32>), compute the average for each pairing
/// of T's.
pub fn average_for_pairs<T: Clone + Eq + Hash>(
    map: &HashMap<(T, T), Vec<f32>>,
) -> HashMap<(T, T), f32> {
    let mut result: HashMap<(T, T), f32> = HashMap::new();
    let len = counts_for_pairs(map);

    for ((a, b), v) in map.iter() {
        if result.get(&(a.clone(), b.clone())) != None {
            continue;
        }

        let len = match len.get(&(a.clone(), b.clone())) {
            Some(v) => v,
            None => len.get(&(b.clone(), a.clone())).unwrap(),
        };

        let mut sum = v.iter().sum::<f32>();
        if a != b {
            sum += match map.get(&(b.clone(), a.clone())) {
                Some(v) => v.iter().sum::<f32>(),
                None => 0.0,
            };
        }

        let avg = sum / (*len as f32);

        result.insert((a.clone(), b.clone()), avg);
    }

    result
}

/// Given a hashmap of ((T, T), Vec<f32>), compute frequency data.
pub fn counts_for_pairs<T: Clone + Eq + Hash, U>(
    map: &HashMap<(T, T), Vec<U>>,
) -> HashMap<(T, T), usize> {
    let mut result = HashMap::new();

    for ((a, b), v) in map.iter() {
        if result.get(&(a.clone(), b.clone())) != None {
            continue;
        }

        let mut count = v.len();
        if a != b {
            count += match map.get(&(b.clone(), a.clone())) {
                Some(v) => v.len(),
                None => 0,
            };
        }

        result.insert((a.clone(), b.clone()), count);
    }

    result
}

/// Pearson moment correlation coefficient. Returns a 3-tuple of (correlation
/// coefficient, *n*).
pub fn pmcc(data: Vec<(f32, f32)>) -> (f32, usize) {
    let x = data.iter().map(|(x, _)| *x);
    let y = data.iter().map(|(_, y)| *y);

    let x_bar = avg(&mut x.clone());
    let y_bar = avg(&mut y.clone());

    let covar: f32 = data.iter().map(|(x, y)| (x - x_bar) * (y - y_bar)).sum();
    let x_var: f32 = x.clone().map(|x| (x - x_bar).powf(2.0)).sum::<f32>().sqrt();
    let y_var: f32 = y.clone().map(|y| (y - y_bar).powf(2.0)).sum::<f32>().sqrt();

    let correlation_coefficient = covar / (x_var * y_var);

    (correlation_coefficient, data.len())
}

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

    #[test]
    fn average_for_simple_pairs() {
        let mut test: HashMap<(u32, u32), Vec<f32>> = HashMap::new();
        test.insert((1, 1), vec![0.0, 1.0]);
        test.insert((1, 2), vec![0.0, 1.0]);
        let test = average_for_pairs(&test);

        let mut expected: HashMap<(u32, u32), f32> = HashMap::new();
        expected.insert((1, 1), 0.5);
        expected.insert((1, 2), 0.5);

        let one_one = test.get(&(1, 1)).unwrap();
        let one_two = test.get(&(1, 2)).unwrap();
        assert!(approx_eq!(f32, *one_one, 0.5));
        assert!(approx_eq!(f32, *one_two, 0.5));
    }

    #[test]
    fn average_for_overlapping_pairs() {
        let mut test: HashMap<(u32, u32), Vec<f32>> = HashMap::new();
        test.insert((2, 1), vec![0.0, 1.0]);
        test.insert((1, 2), vec![0.0, 1.0]);
        let test = average_for_pairs(&test);

        let mut expected: HashMap<(u32, u32), f32> = HashMap::new();
        expected.insert((2, 1), 0.5);

        let one_two = test.get(&(2, 1)).unwrap();
        println!("{}\n", *one_two);
        assert!(approx_eq!(f32, *one_two, 0.5));
    }

    #[test]
    fn counts_for_simple_pairs() {
        let mut test: HashMap<(u32, u32), Vec<f32>> = HashMap::new();
        test.insert((1, 1), vec![0.0, 1.0]);
        test.insert((1, 2), vec![0.0, 1.0]);
        let test = counts_for_pairs(&test);

        assert_eq!(*(test.get(&(1, 1)).unwrap()), 2 as usize);
    }

    #[test]
    fn counts_for_overlapping_pairs() {
        let mut test: HashMap<(u32, u32), Vec<f32>> = HashMap::new();
        test.insert((2, 1), vec![0.0, 1.0]);
        test.insert((1, 2), vec![0.0, 1.0]);
        let test = counts_for_pairs(&test);

        assert_eq!(*(test.get(&(2, 1)).unwrap()), 4 as usize);
    }
}