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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579
use crate::*; //{RError, nodata_error, sumn, fromop, Params, MutVecg, Stats, Vecg, RE};
use indxvec::Vecops;
use medians::{Medianf64,Median};
use core::cmp::Ordering::*;
impl<T> Stats for &[T]
where
T: Clone + PartialOrd + Into<f64>
{
/// Vector magnitude
fn vmag(self) -> f64 {
match self.len() {
0 => 0_f64,
1 => self[0].clone().into(),
_ => self.iter().map(|x| x.clone().into().powi(2)).sum::<f64>().sqrt(),
}
}
/// Vector magnitude squared (sum of squares)
fn vmagsq(self) -> f64 {
match self.len() {
0 => 0_f64,
1 => self[0].clone().into().powi(2),
_ => self.iter().map(|x| x.clone().into().powi(2)).sum::<f64>(),
}
}
/// Vector with reciprocal components
fn vreciprocal(self) -> Result<Vec<f64>, RE> {
if self.is_empty() {
return nodata_error("vreciprocal: empty self vec");
};
self.iter()
.map(|component| -> Result<f64, RE> {
let c: f64 = component.clone().into();
if c.is_normal() {
Ok(1.0 / c)
} else {
arith_error(format!("vreciprocal: bad component {c}"))
}
})
.collect::<Result<Vec<f64>, RE>>()
}
/// Vector with inverse magnitude
fn vinverse(self) -> Result<Vec<f64>, RE> {
if self.is_empty() {
return nodata_error("vinverse: empty self vec");
};
let vmagsq = self.vmagsq();
if vmagsq > 0.0 {
Ok(self.iter().map(|x| x.clone().into() / vmagsq).collect())
} else {
data_error("vinverse: can not invert zero vector")
}
}
// Negated vector (all components swap sign)
fn negv(self) -> Result<Vec<f64>, RE> {
if self.is_empty() {
return nodata_error("negv: empty self vec");
};
Ok(self.iter().map(|x| (-x.clone().into())).collect())
}
/// Unit vector
fn vunit(self) -> Result<Vec<f64>, RE> {
if self.is_empty() {
return nodata_error("vunit: empty self vec");
};
let mag = self.vmag();
if mag > 0.0 {
Ok(self.iter().map(|x| x.clone().into() / mag).collect())
} else {
data_error("vunit: can not make zero vector into a unit vector")
}
}
/// Harmonic spread from median
fn hmad(self) -> Result<f64, RE> {
let n = self.len();
if n == 0 {
return nodata_error("hmad: empty self");
};
let fself = self.iter().map(|x| x.clone().into()).collect::<Vec<f64>>();
let recmedian = 1.0 / fself.medf_checked()?;
let recmad = self
.iter()
.map(|x| -> Result<f64, RE> {
let fx: f64 = x.clone().into();
if !fx.is_normal() {
return arith_error("hmad: attempt to divide by zero");
};
Ok((recmedian - 1.0 / fx).abs())
})
.collect::<Result<Vec<f64>, RE>>()?
.medf_unchecked();
Ok(recmedian / recmad)
}
/// Arithmetic mean
/// # Example
/// ```
/// use rstats::Stats;
/// let v1 = vec![1_f64,2.,3.,4.,5.,6.,7.,8.,9.,10.,11.,12.,13.,14.];
/// assert_eq!(v1.amean().unwrap(),7.5_f64);
/// ```
fn amean(self) -> Result<f64, RE> {
let n = self.len();
if n > 0 {
Ok(self.iter().map(|x| x.clone().into()).sum::<f64>() / (n as f64))
} else {
nodata_error("amean: empty self vec")
}
}
/// Median and Mad packed into in Params struct
fn medmad(self) -> Result<Params, RE> {
let median = self.qmedian_by(&mut |a,b| a
.partial_cmp(b).unwrap_or(Equal), fromop)?;
Ok(Params{centre:median,spread:self.mad(median, fromop)})
}
/// Arithmetic mean and (population) standard deviation
/// # Example
/// ```
/// use rstats::Stats;
/// let v1 = vec![1_f64,2.,3.,4.,5.,6.,7.,8.,9.,10.,11.,12.,13.,14.];
/// let res = v1.ameanstd().unwrap();
/// assert_eq!(res.centre,7.5_f64);
/// assert_eq!(res.spread,4.031128874149275_f64);
/// ```
fn ameanstd(self) -> Result<Params, RE> {
let n = self.len();
if n == 0 {
return Err(RError::NoDataError("empty self vec".to_owned()));
};
let nf = n as f64;
let mut sx2 = 0_f64;
let mean = self
.iter()
.map(|x| {
let fx: f64 = x.clone().into();
sx2 += fx * fx;
fx
})
.sum::<f64>()
/ nf;
Ok(Params {
centre: mean,
spread: (sx2 / nf - mean.powi(2)).sqrt(),
})
}
/// Linearly weighted arithmetic mean of an f64 slice.
/// Linearly ascending weights from 1 to n.
/// Time dependent data should be in the order of time increasing.
/// Then the most recent gets the most weight.
/// # Example
/// ```
/// use rstats::Stats;
/// let v1 = vec![1_f64,2.,3.,4.,5.,6.,7.,8.,9.,10.,11.,12.,13.,14.];
/// assert_eq!(v1.awmean().unwrap(),9.666666666666666_f64);
/// ```
fn awmean(self) -> Result<f64, RE> {
let n = self.len();
if n == 0 {
return Err(RError::NoDataError("empty self vec".to_owned()));
};
let mut iw = 0_f64; // descending linear weights
Ok(self
.iter()
.map(|x| {
iw += 1_f64;
iw * x.clone().into()
})
.sum::<f64>()
/ (sumn(n) as f64))
}
/// Linearly weighted arithmetic mean and standard deviation of an f64 slice.
/// Linearly ascending weights from 1 to n.
/// Time dependent data should be in the order of time increasing.
/// # Example
/// ```
/// use rstats::Stats;
/// let v1 = vec![1_f64,2.,3.,4.,5.,6.,7.,8.,9.,10.,11.,12.,13.,14.];
/// let res = v1.awmeanstd().unwrap();
/// assert_eq!(res.centre,9.666666666666666_f64);
/// assert_eq!(res.spread,3.399346342395192_f64);
/// ```
fn awmeanstd(self) -> Result<Params, RE> {
let n = self.len();
if n == 0 {
return Err(RError::NoDataError("empty self vec".to_owned()));
};
let mut sx2 = 0_f64;
let mut w = 0_f64; // descending linear weights
let nf = sumn(n) as f64;
let centre = self
.iter()
.map(|x| {
let fx: f64 = x.clone().into();
w += 1_f64;
let wx = w * fx;
sx2 += wx * fx;
wx
})
.sum::<f64>()
/ nf;
Ok(Params {
centre,
spread: (sx2 / nf - centre.powi(2)).sqrt(),
})
}
/// Harmonic mean of an f64 slice.
/// # Example
/// ```
/// use rstats::Stats;
/// let v1 = vec![1_f64,2.,3.,4.,5.,6.,7.,8.,9.,10.,11.,12.,13.,14.];
/// assert_eq!(v1.hmean().unwrap(),4.305622526633627_f64);
/// ```
fn hmean(self) -> Result<f64, RE> {
let n = self.len();
if n == 0 {
return Err(RError::NoDataError("empty self vec".to_owned()));
};
let mut sum = 0_f64;
for x in self {
let fx: f64 = x.clone().into();
if !fx.is_normal() {
return Err(RError::ArithError("attempt to divide by zero".to_owned()));
};
sum += 1.0 / fx
}
Ok(n as f64 / sum)
}
/// Harmonic mean and standard deviation
/// std is based on reciprocal moments
/// # Example
/// ```
/// use rstats::Stats;
/// let v1 = vec![1_f64,2.,3.,4.,5.,6.,7.,8.,9.,10.,11.,12.,13.,14.];
/// let res = v1.hmeanstd().unwrap();
/// assert_eq!(res.centre,4.305622526633627_f64);
/// assert_eq!(res.spread,1.1996764516690959_f64);
/// ```
fn hmeanstd(self) -> Result<Params, RE> {
let n = self.len();
if n == 0 {
return Err(RError::NoDataError("empty self vec".to_owned()));
};
let nf = n as f64;
let mut sx2 = 0_f64;
let mut sx = 0_f64;
for x in self {
let fx: f64 = x.clone().into();
if !fx.is_normal() {
return Err(RError::ArithError("attempt to divide by zero".to_owned()));
};
let rx = 1_f64 / fx; // work with reciprocals
sx2 += rx * rx;
sx += rx;
}
let recipmean = sx / nf;
Ok(Params {
centre: 1.0 / recipmean,
spread: ((sx2 / nf - recipmean.powi(2)) / (recipmean.powi(4)) / nf).sqrt(),
})
}
/// Linearly weighted harmonic mean of an f64 slice.
/// Linearly ascending weights from 1 to n.
/// Time dependent data should be ordered by increasing time.
/// # Example
/// ```
/// use rstats::Stats;
/// let v1 = vec![1_f64,2.,3.,4.,5.,6.,7.,8.,9.,10.,11.,12.,13.,14.];
/// assert_eq!(v1.hwmean().unwrap(),7.5_f64);
/// ```
fn hwmean(self) -> Result<f64, RE> {
let n = self.len();
if n == 0 {
return Err(RError::NoDataError("empty self vec".to_owned()));
};
let mut sum = 0_f64;
let mut w = 0_f64;
for x in self {
let fx: f64 = x.clone().into();
if !fx.is_normal() {
return Err(RError::ArithError("attempt to divide by zero".to_owned()));
};
w += 1_f64;
sum += w / fx;
}
Ok(sumn(n) as f64 / sum) // reciprocal of the mean of reciprocals
}
/// Weighted harmonic mean and standard deviation
/// std is based on reciprocal moments
/// # Example
/// ```
/// use rstats::Stats;
/// let v1 = vec![1_f64,2.,3.,4.,5.,6.,7.,8.,9.,10.,11.,12.,13.,14.];
/// let res = v1.hmeanstd().unwrap();
/// assert_eq!(res.centre,4.305622526633627_f64);
/// assert_eq!(res.spread,1.1996764516690959_f64);
/// ```
fn hwmeanstd(self) -> Result<Params, RE> {
let n = self.len();
if n == 0 {
return Err(RError::NoDataError("empty self vec".to_owned()));
};
let nf = sumn(n) as f64;
let mut sx2 = 0_f64;
let mut sx = 0_f64;
let mut w = 0_f64;
for x in self {
w += 1_f64;
let fx: f64 = x.clone().into();
if !fx.is_normal() {
return Err(RError::ArithError("attempt to divide by zero".to_owned()));
};
sx += w / fx; // work with reciprocals
sx2 += w / (fx * fx);
}
let recipmean = sx / nf;
Ok(Params {
centre: 1.0 / recipmean,
spread: ((sx2 / nf - recipmean.powi(2)) / (recipmean.powi(4)) / nf).sqrt(),
})
}
/// Geometric mean of an i64 slice.
/// The geometric mean is just an exponential of an arithmetic mean
/// of log data (natural logarithms of the data items).
/// The geometric mean is less sensitive to outliers near maximal value.
/// Zero valued data is not allowed!
/// # Example
/// ```
/// use rstats::Stats;
/// let v1 = vec![1_f64,2.,3.,4.,5.,6.,7.,8.,9.,10.,11.,12.,13.,14.];
/// assert_eq!(v1.gmean().unwrap(),6.045855171418503_f64);
/// ```
fn gmean(self) -> Result<f64, RE> {
let n = self.len();
if n == 0 {
return Err(RError::NoDataError("empty self vec".to_owned()));
};
let mut sum = 0_f64;
for x in self {
let fx: f64 = x.clone().into();
if !fx.is_normal() {
return Err(RError::ArithError(
"gmean attempt to take ln of zero".to_owned(),
));
};
sum += fx.ln()
}
Ok((sum / (n as f64)).exp())
}
/// Geometric mean and std ratio of an f64 slice.
/// Zero valued data is not allowed.
/// Std of ln data becomes a ratio after conversion back.
/// # Example
/// ```
/// use rstats::Stats;
/// let v1 = vec![1_f64,2.,3.,4.,5.,6.,7.,8.,9.,10.,11.,12.,13.,14.];
/// let res = v1.gmeanstd().unwrap();
/// assert_eq!(res.centre,6.045855171418503_f64);
/// assert_eq!(res.spread,2.1084348239406303_f64);
/// ```
fn gmeanstd(self) -> Result<Params, RE> {
let n = self.len();
if n == 0 {
return Err(RError::NoDataError("empty self vec".to_owned()));
};
let mut sum = 0_f64;
let mut sx2 = 0_f64;
for x in self {
let fx: f64 = x.clone().into();
if !fx.is_normal() {
return Err(RError::ArithError(
"gmeanstd attempt to take ln of zero".to_owned(),
));
};
let lx = fx.ln();
sum += lx;
sx2 += lx * lx
}
sum /= n as f64;
Ok(Params {
centre: sum.exp(),
spread: (sx2 / (n as f64) - sum.powi(2)).sqrt().exp(),
})
}
/// Linearly weighted geometric mean of an i64 slice.
/// Ascending weights from 1 down to n.
/// Time dependent data should be in time increasing order.
/// The geometric mean is an exponential of an arithmetic mean
/// of log data (natural logarithms of the data items).
/// The geometric mean is less sensitive to outliers near maximal value.
/// Zero valued data is not allowed!
/// # Example
/// ```
/// use rstats::Stats;
/// let v1 = vec![1_f64,2.,3.,4.,5.,6.,7.,8.,9.,10.,11.,12.,13.,14.];
/// assert_eq!(v1.gwmean().unwrap(),8.8185222496341_f64);
/// ```
fn gwmean(self) -> Result<f64, RE> {
let n = self.len();
if n == 0 {
return Err(RError::NoDataError("empty self vec".to_owned()));
};
let mut w = 0_f64; // ascending weights
let mut sum = 0_f64;
for x in self {
let fx: f64 = x.clone().into();
if !fx.is_normal() {
return Err(RError::ArithError(
"gwmean attempt to take ln of zero".to_owned(),
));
};
w += 1_f64;
sum += w * fx.ln();
}
Ok((sum / sumn(n) as f64).exp())
}
/// Linearly weighted version of gmeanstd.
/// # Example
/// ```
/// use rstats::Stats;
/// let v1 = vec![1_f64,2.,3.,4.,5.,6.,7.,8.,9.,10.,11.,12.,13.,14.];
/// let res = v1.gwmeanstd().unwrap();
/// assert_eq!(res.centre,8.8185222496341_f64);
/// assert_eq!(res.spread,1.626825493266009_f64);
/// ```
fn gwmeanstd(self) -> Result<Params, RE> {
let n = self.len();
if n == 0 {
return Err(RError::NoDataError("empty self vec".to_owned()));
};
let mut w = 0_f64; // ascending weights
let mut sum = 0_f64;
let mut sx2 = 0_f64;
for x in self {
let fx: f64 = x.clone().into();
if !fx.is_normal() {
return Err(RError::ArithError(
"gwmeanstd attempt to take ln of zero".to_owned(),
));
};
let lnx = fx.ln();
w += 1_f64;
sum += w * lnx;
sx2 += w * lnx * lnx;
}
let nf = sumn(n) as f64;
sum /= nf;
Ok(Params {
centre: sum.exp(),
spread: (sx2 / nf - sum.powi(2)).sqrt().exp(),
})
}
/// Probability density function of a sorted slice with repeats.
/// Repeats are counted and removed
fn pdf(self) -> Vec<f64> {
let nf = self.len() as f64;
let mut res: Vec<f64> = Vec::new();
let mut count = 1_usize; // running count
let mut lastval = &self[0];
self.iter().skip(1).for_each(|s| {
if *s > *lastval {
// new value encountered
res.push((count as f64) / nf); // save previous probability
lastval = s; // new value
count = 1_usize; // reset counter
} else {
count += 1;
}
});
res.push((count as f64) / nf); // flush the rest!
res
}
/// Information (entropy) (in nats)
fn entropy(self) -> f64 {
let pdfv = self.sortm(true).pdf();
pdfv.iter().map(|&x| -x * (x.ln())).sum()
}
/// (Auto)correlation coefficient of pairs of successive values of (time series) f64 variable.
/// # Example
/// ```
/// use rstats::Stats;
/// let v1 = vec![1_f64,2.,3.,4.,5.,6.,7.,8.,9.,10.,11.,12.,13.,14.];
/// assert_eq!(v1.autocorr().unwrap(),0.9984603532054123_f64);
/// ```
fn autocorr(self) -> Result<f64, RE> {
let (mut sx, mut sy, mut sxy, mut sx2, mut sy2) = (0_f64, 0_f64, 0_f64, 0_f64, 0_f64);
let n = self.len();
if n < 2 {
return Err(RError::NoDataError(
"autocorr needs a Vec of at least two items".to_owned(),
));
};
let mut x: f64 = self[0].clone().into();
self.iter().skip(1).for_each(|si| {
let y: f64 = si.clone().into();
sx += x;
sy += y;
sxy += x * y;
sx2 += x * x;
sy2 += y * y;
x = y
});
let nf = n as f64;
Ok((sxy - sx / nf * sy) / ((sx2 - sx / nf * sx) * (sy2 - sy / nf * sy)).sqrt())
}
/// Linear transform to interval [0,1]
fn lintrans(self) -> Result<Vec<f64>, RE> {
let mm = self.minmax();
let min = mm.min.into();
let range = mm.max.into() - min;
if range == 0_f64 {
return Err(RError::ArithError(
"lintrans self has zero range".to_owned(),
));
};
Ok(self
.iter()
.map(|x| (x.clone().into() - min) / range)
.collect())
}
/// Linearly weighted approximate time series derivative at the last point (present time).
/// Weighted sum (backwards half filter), minus the median.
/// Rising values return positive result and vice versa.
fn dfdt(self, centre: f64) -> Result<f64, RE> {
let len = self.len();
if len < 2 {
return Err(RError::NoDataError(format!(
"dfdt time series too short: {len}"
)));
};
let mut weight = 0_f64;
let mut sumwx = 0_f64;
for x in self.iter() {
weight += 1_f64;
sumwx += weight * x.clone().into();
}
Ok(sumwx / (sumn(len) as f64) - centre)
}
/// Householder reflector
fn house_reflector(self) -> Vec<f64> {
let norm = self.vmag();
if norm.is_normal() {
let mut u = self.smult(1. / norm);
if u[0] < 0. {
u[0] -= 1.;
} else {
u[0] += 1.;
};
let uzero = 1.0 / (u[0].abs().sqrt());
u.mutsmult(uzero);
return u;
};
let mut u = vec![0.; self.len()];
u[0] = std::f64::consts::SQRT_2;
u
}
}