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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281
use num_traits::ToPrimitive;
#[cfg(feature = "serde1")]
use serde_derive::{Deserialize, Serialize};
use crate::Merge;
/// Estimate the arithmetic means and the covariance of a sequence of number pairs
/// ("population").
///
/// Because the variances are calculated as well, this can be used to calculate the Pearson
/// correlation coefficient.
///
///
/// ## Example
///
/// ```
/// use average::Covariance;
///
/// let a: Covariance = [(1., 5.), (2., 4.), (3., 3.), (4., 2.), (5., 1.)].iter().cloned().collect();
/// assert_eq!(a.mean_x(), 3.);
/// assert_eq!(a.mean_y(), 3.);
/// assert_eq!(a.population_covariance(), -2.0);
/// assert_eq!(a.sample_covariance(), -2.5);
/// ```
#[derive(Debug, Clone)]
#[cfg_attr(feature = "serde1", derive(Serialize, Deserialize))]
pub struct Covariance {
avg_x: f64,
sum_x_2: f64,
avg_y: f64,
sum_y_2: f64,
sum_prod: f64,
n: u64,
}
impl Covariance {
/// Create a new covariance estimator.
#[inline]
pub fn new() -> Covariance {
Covariance {
avg_x: 0.,
sum_x_2: 0.,
avg_y: 0.,
sum_y_2: 0.,
sum_prod: 0.,
n: 0,
}
}
/// Add an observation sampled from the population.
#[inline]
pub fn add(&mut self, x: f64, y: f64) {
self.n += 1;
let n = self.n.to_f64().unwrap();
let delta_x = x - self.avg_x;
let delta_x_n = delta_x / n;
let delta_y_n = (y - self.avg_y) / n;
self.avg_x += delta_x_n;
self.sum_x_2 += delta_x_n * delta_x_n * n * (n - 1.);
self.avg_y += delta_y_n;
self.sum_y_2 += delta_y_n * delta_y_n * n * (n - 1.);
self.sum_prod += delta_x * (y - self.avg_y);
}
/// Calculate the population covariance of the sample.
///
/// This is a biased estimator of the covariance of the population.
///
/// Returns NaN for an empty sample.
#[inline]
pub fn population_covariance(&self) -> f64 {
if self.n < 1 {
return f64::NAN;
}
self.sum_prod / self.n.to_f64().unwrap()
}
/// Calculate the sample covariance.
///
/// This is an unbiased estimator of the covariance of the population.
///
/// Returns NaN for samples of size 1 or less.
#[inline]
pub fn sample_covariance(&self) -> f64 {
if self.n < 2 {
return f64::NAN;
}
self.sum_prod / (self.n - 1).to_f64().unwrap()
}
/// Calculate the population Pearson correlation coefficient.
///
/// Returns NaN for an empty sample.
#[cfg(any(feature = "std", feature = "libm"))]
#[cfg_attr(doc_cfg, doc(cfg(any(feature = "std", feature = "libm"))))]
#[inline]
pub fn pearson(&self) -> f64 {
if self.n < 2 {
return f64::NAN;
}
self.sum_prod / num_traits::Float::sqrt(self.sum_x_2 * self.sum_y_2)
}
/// Return the sample size.
#[inline]
pub fn len(&self) -> u64 {
self.n
}
/// Determine whether the sample is empty.
#[inline]
pub fn is_empty(&self) -> bool {
self.n == 0
}
/// Estimate the mean of the `x` population.
///
/// Returns NaN for an empty sample.
#[inline]
pub fn mean_x(&self) -> f64 {
if self.n > 0 { self.avg_x } else { f64::NAN }
}
/// Estimate the mean of the `y` population.
///
/// Returns NaN for an empty sample.
#[inline]
pub fn mean_y(&self) -> f64 {
if self.n > 0 { self.avg_y } else { f64::NAN }
}
/// Calculate the sample variance of `x`.
///
/// This is an unbiased estimator of the variance of the population.
///
/// Returns NaN for samples of size 1 or less.
#[inline]
pub fn sample_variance_x(&self) -> f64 {
if self.n < 2 {
return f64::NAN;
}
self.sum_x_2 / (self.n - 1).to_f64().unwrap()
}
/// Calculate the population variance of the sample for `x`.
///
/// This is a biased estimator of the variance of the population.
///
/// Returns NaN for an empty sample.
#[inline]
pub fn population_variance_x(&self) -> f64 {
if self.n == 0 {
return f64::NAN;
}
self.sum_x_2 / self.n.to_f64().unwrap()
}
/// Calculate the sample variance of `y`.
///
/// This is an unbiased estimator of the variance of the population.
///
/// Returns NaN for samples of size 1 or less.
#[inline]
pub fn sample_variance_y(&self) -> f64 {
if self.n < 2 {
return f64::NAN;
}
self.sum_y_2 / (self.n - 1).to_f64().unwrap()
}
/// Calculate the population variance of the sample for `y`.
///
/// This is a biased estimator of the variance of the population.
///
/// Returns NaN for an empty sample.
#[inline]
pub fn population_variance_y(&self) -> f64 {
if self.n == 0 {
return f64::NAN;
}
self.sum_y_2 / self.n.to_f64().unwrap()
}
// TODO: Standard deviation and standard error
}
impl core::default::Default for Covariance {
fn default() -> Covariance {
Covariance::new()
}
}
impl Merge for Covariance {
/// Merge another sample into this one.
///
/// ## Example
///
/// ```
/// use average::{Covariance, Merge};
///
/// let sequence: &[(f64, f64)] = &[(1., 2.), (3., 4.), (5., 6.), (7., 8.), (9., 10.)];
/// let (left, right) = sequence.split_at(3);
/// let cov_total: Covariance = sequence.iter().collect();
/// let mut cov_left: Covariance = left.iter().collect();
/// let cov_right: Covariance = right.iter().collect();
/// cov_left.merge(&cov_right);
/// assert_eq!(cov_total.population_covariance(), cov_left.population_covariance());
/// ```
#[inline]
fn merge(&mut self, other: &Covariance) {
if other.n == 0 {
return;
}
if self.n == 0 {
*self = other.clone();
return;
}
let delta_x = other.avg_x - self.avg_x;
let delta_y = other.avg_y - self.avg_y;
let len_self = self.n.to_f64().unwrap();
let len_other = other.n.to_f64().unwrap();
let len_total = len_self + len_other;
self.avg_x = (len_self * self.avg_x + len_other * other.avg_x) / len_total;
self.sum_x_2 += other.sum_x_2 + delta_x*delta_x * len_self * len_other / len_total;
self.avg_y = (len_self * self.avg_y + len_other * other.avg_y) / len_total;
self.sum_y_2 += other.sum_y_2 + delta_y*delta_y * len_self * len_other / len_total;
self.sum_prod += other.sum_prod + delta_x*delta_y * len_self * len_other / len_total;
self.n += other.n;
}
}
impl core::iter::FromIterator<(f64, f64)> for Covariance {
fn from_iter<T>(iter: T) -> Covariance
where
T: IntoIterator<Item = (f64, f64)>,
{
let mut cov = Covariance::new();
for (x, y) in iter {
cov.add(x, y);
}
cov
}
}
impl core::iter::Extend<(f64, f64)> for Covariance {
fn extend<T: IntoIterator<Item = (f64, f64)>>(&mut self, iter: T) {
for (x, y) in iter {
self.add(x, y);
}
}
}
impl<'a> core::iter::FromIterator<&'a (f64, f64)> for Covariance {
fn from_iter<T>(iter: T) -> Covariance
where
T: IntoIterator<Item = &'a (f64, f64)>,
{
let mut cov = Covariance::new();
for &(x, y) in iter {
cov.add(x, y);
}
cov
}
}
impl<'a> core::iter::Extend<&'a (f64, f64)> for Covariance {
fn extend<T: IntoIterator<Item = &'a (f64, f64)>>(&mut self, iter: T) {
for &(x, y) in iter {
self.add(x, y);
}
}
}