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

//! Bivariate analysis

mod bootstrap;
pub mod regression;
mod resamples;

use crate::stats::bivariate::resamples::Resamples;
use crate::stats::float::Float;
use crate::stats::tuple::{Tuple, TupledDistributionsBuilder};
use crate::stats::univariate::Sample;
use rayon::iter::{IntoParallelIterator, ParallelIterator};

/// Bivariate `(X, Y)` data
///
/// Invariants:
///
/// - No `NaN`s in the data
/// - At least two data points in the set
pub struct Data<'a, X, Y>(&'a [X], &'a [Y]);

impl<'a, X, Y> Copy for Data<'a, X, Y> {}

#[cfg_attr(feature = "cargo-clippy", allow(clippy::expl_impl_clone_on_copy))]
impl<'a, X, Y> Clone for Data<'a, X, Y> {
    fn clone(&self) -> Data<'a, X, Y> {
        *self
    }
}

impl<'a, X, Y> Data<'a, X, Y> {
    /// Returns the length of the data set
    pub fn len(&self) -> usize {
        self.0.len()
    }

    /// Iterate over the data set
    pub fn iter(&self) -> Pairs<'a, X, Y> {
        Pairs {
            data: *self,
            state: 0,
        }
    }
}

impl<'a, X, Y> Data<'a, X, Y>
where
    X: Float,
    Y: Float,
{
    /// Creates a new data set from two existing slices
    pub fn new(xs: &'a [X], ys: &'a [Y]) -> Data<'a, X, Y> {
        assert!(
            xs.len() == ys.len()
                && xs.len() > 1
                && xs.iter().all(|x| !x.is_nan())
                && ys.iter().all(|y| !y.is_nan())
        );

        Data(xs, ys)
    }

    // TODO Remove the `T` parameter in favor of `S::Output`
    /// Returns the bootstrap distributions of the parameters estimated by the `statistic`
    ///
    /// - Multi-threaded
    /// - Time: `O(nresamples)`
    /// - Memory: `O(nresamples)`
    pub fn bootstrap<T, S>(&self, nresamples: usize, statistic: S) -> T::Distributions
    where
        S: Fn(Data<X, Y>) -> T + Sync,
        T: Tuple + Send,
        T::Distributions: Send,
        T::Builder: Send,
    {
        (0..nresamples)
            .into_par_iter()
            .map_init(
                || Resamples::new(*self),
                |resamples, _| statistic(resamples.next()),
            )
            .fold(
                || T::Builder::new(0),
                |mut sub_distributions, sample| {
                    sub_distributions.push(sample);
                    sub_distributions
                },
            )
            .reduce(
                || T::Builder::new(0),
                |mut a, mut b| {
                    a.extend(&mut b);
                    a
                },
            )
            .complete()
    }

    /// Returns a view into the `X` data
    pub fn x(&self) -> &'a Sample<X> {
        Sample::new(&self.0)
    }

    /// Returns a view into the `Y` data
    pub fn y(&self) -> &'a Sample<Y> {
        Sample::new(&self.1)
    }
}

/// Iterator over `Data`
pub struct Pairs<'a, X: 'a, Y: 'a> {
    data: Data<'a, X, Y>,
    state: usize,
}

impl<'a, X, Y> Iterator for Pairs<'a, X, Y> {
    type Item = (&'a X, &'a Y);

    fn next(&mut self) -> Option<(&'a X, &'a Y)> {
        if self.state < self.data.len() {
            let i = self.state;
            self.state += 1;

            // This is safe because i will always be < self.data.{0,1}.len()
            debug_assert!(i < self.data.0.len());
            debug_assert!(i < self.data.1.len());
            unsafe { Some((self.data.0.get_unchecked(i), self.data.1.get_unchecked(i))) }
        } else {
            None
        }
    }
}