# Asymmetric Kiener Distribution K7 (K2) Density, distribution function, quantile function, random generation, value-at-risk, expected shortfall (+ signed left/right tail mean) and additional formulae for asymmetric Kiener distribution K7 = K2. With K7, the vector of parameters is provided as `coefk`, usually estimated with `paramkienerX` (and ~X5,~X7) or `regkienerLX$coefk`. Main inputs can be supplied as vector (`x,q,p`) and matrix (`coefk`) and the resulting output is a matrix (useful for simulation). ```r dkiener7(x, coefk = c(0, 1, 3.2, 3.2, 3.2, 0, 0), log = FALSE) pkiener7(q, coefk = c(0, 1, 3.2, 3.2, 3.2, 0, 0), lower.tail = TRUE, log.p = FALSE) qkiener7(p, coefk = c(0, 1, 3.2, 3.2, 3.2, 0, 0), lower.tail = TRUE, log.p = FALSE) rkiener7(n, coefk = c(0, 1, 3.2, 3.2, 3.2, 0, 0), same_p = FALSE) dpkiener7(p, coefk = c(0, 1, 3.2, 3.2, 3.2, 0, 0), log = FALSE) dqkiener7(p, coefk = c(0, 1, 3.2, 3.2, 3.2, 0, 0), log = FALSE) lkiener7(x, coefk = c(0, 1, 3.2, 3.2, 3.2, 0, 0)) dlkiener7(lp, coefk = c(0, 1, 3.2, 3.2, 3.2, 0, 0), log = FALSE) qlkiener7(lp, coefk = c(0, 1, 3.2, 3.2, 3.2, 0, 0), lower.tail = TRUE) varkiener7(p, coefk = c(0, 1, 3.2, 3.2, 3.2, 0, 0), lower.tail = TRUE, log.p = FALSE) ltmkiener7(p, coefk = c(0, 1, 3.2, 3.2, 3.2, 0, 0), lower.tail = TRUE, log.p = FALSE) rtmkiener7(p, coefk = c(0, 1, 3.2, 3.2, 3.2, 0, 0), lower.tail = TRUE, log.p = FALSE) dtmqkiener7(p, coefk = c(0, 1, 3.2, 3.2, 3.2, 0, 0), lower.tail = TRUE, log.p = FALSE) eskiener7(p, coefk = c(0, 1, 3.2, 3.2, 3.2, 0, 0), lower.tail = TRUE, log.p = FALSE, signedES = FALSE) ``` ## Arguments - `x`: vector of quantiles. - `coefk`: vector of 7 parameters `c(m,g,a,k,w,d,e)` or matrix with 7 columns. - `log`: logical. If TRUE, densities are given in log scale. - `q`: vector of quantiles. - `lower.tail`: logical. If TRUE, use p. If FALSE, use 1-p. - `log.p`: logical. If TRUE, probabilities p are given as log(p). - `p`: vector of probabilities. - `n`: integer. Number of observations. If length(n) > 1, the length is taken to be the number required. - `same_p`: logical. If FALSE (default), random probabilies are generated on the fly. If TRUE, the same set of random probabilities is used for each line of coefk (if coefk is a matrix). - `lp`: vector of logit of probabilities. - `signedES`: logical. FALSE (default) returns positive numbers for left and right tails. TRUE returns negative number (= `ltmkiener7`) for left tail and positive number (= `rtmkiener7`) for right tail. ## Details Kiener distributions use the following parameters, some of them being redundant. See `aw2k` and `pk2pk` for the formulas and the conversion between parameters: * `m` (mu) is the median of the distribution. * `g` (gamma) is the scale parameter. * `a` (alpha) is the left tail parameter. * `k` (kappa) is the harmonic mean of `a` and `w` and describes a global tail parameter. * `w` (omega) is the right tail parameter. * `d` (delta) is the distortion parameter. * `e` (epsilon) is the eccentricity parameter. Kiener distribution `K7` is designed after `kiener2` but uses as input `coefk` rather than `m`, `g`, `a` and `w`. `m` is the median of the distribution. `g` is the scale parameter and is linked for any value of `a` and `w` to the density at the median through the relation $$ g * dkiener7(x=m, coefk=coefk) = \frac{\pi}{4\sqrt{3}} \approx 0.453%g * dkiener7(x=m, coefk=coefk) = pi/4/sqrt(3) = 0.453 approximatively $$ When `a = Inf` and `w = Inf`, `g` is very close to `sd(x)`. NOTE: In order to match this standard deviation, the value of `g` has been updated from versions < 1.9.0 by a factor $ 2 pi/sqrt(3) $. The functions `dkiener2347`, `pkiener2347` and `lkiener2347` have no explicit forms. Due to a poor optimization algorithm, their calculations in versions < 1.9 were unreliable. In versions > 1.9, a much better algorithm was found and the optimization is conducted in a fast way to avoid a lengthy optimization. The two extreme elements (minimum, maximum) of the given `x` or `q` arguments are sent to a second order optimizer that minimize the residual error of the `lkiener2347` function and return the estimated lower and upper logit values. Then a sequence of logit values of length 51 times the length of `x` or `q` is generated between these lower and upper values and the corresponding quantiles are calculated with the function `qlkiener2347`. These 51 times more numerous quantiles are then compared to the original `x` or `q` arguments and the closest values with their associated logit values are selected. The probabilities are then calculated with the function `invlogit` and the densities are calculated with the function `dlkiener2347`. The accuracy of this approach depends on the sparsity of the initial `x` or `q` sequences. A 4 digits accuracy can be expected, enough for most usages. `qkiener7` function is defined for p in (0, 1) by: $$ qkiener7(p, coefk) = m + \frac{\sqrt{3}}{\pi}*g*k*\left(-exp\left(-\frac{logit(p)}{a} +\frac{logit(p)}{w}\right)\right)%qkiener7(p, coefk) = m + sqrt(3)/pi*g*k*(-exp(-logit(p)/a) +exp(logit(p)/w)) $$ where k is the harmonic mean of the tail parameters `a` and `w` calculated by $k = aw2k(a, w)$. `rkiener7` generates `n` random quantiles. In addition to the classical d, p, q, r functions, the prefixes dp, dq, l, dl, ql are also provided. `dpkiener7` is the density function calculated from the probability p. It is defined for p in (0, 1) by: $$ dpkiener7(p, coefk) = \frac{\pi}{\sqrt{3}}\frac{p(1-p)}{g}\frac{2}{k}\left[ +\frac{1}{a}exp\left(-\frac{logit(p)}{a}\right)+\frac{1}{w}exp\left( \frac{logit(p)}{w}\right) \right]^{-1}%dpkiener7(p, coefk) = pi/sqrt(3)*p*(1-p)/g*2/k/(+exp(-logit(p)/a)/a +exp(logit(p)/w)/w) $$ `dqkiener7` is the derivate of the quantile function calculated from the probability p. It is defined for p in (0, 1) by: $$ dqkiener7(p, coefk) = \frac{\sqrt{3}}{\pi}\frac{g}{p(1-p)}\frac{k}{2}\left[ +\frac{1}{a}exp\left(-\frac{logit(p)}{a}\right)+\frac{1}{w}exp\left( \frac{logit(p)}{w}\right) \right]%dqkiener4(p,m,g,k,e) = sqrt(3)/pi*g/p/(1-p)*k/2*( exp(-logit(p)/a)/a + exp(logit(p)/w)/w) $$ with `a` and `w` extracted from `coefk`. `dlkiener7` is the density function calculated from the logit of the probability lp = logit(p) defined in (-Inf, +Inf). The formula is adapted from distribution K2: $$ dlkiener7(lp, coefk) = \frac{\pi}{\sqrt{3}}\frac{p(1-p)}{g}\frac{2}{k}\left[ +\frac{1}{a}exp\left(-\frac{lp}{a}\right)+\frac{1}{w}exp\left( \frac{lp}{w}\right) \right]^{-1}%dlkiener7(lp, coefk) = pi/sqrt(3)*p*(1-p)/g*2/k/(+exp(-lp/a)/a +exp(lp/w)/w) $$ `qlkiener7` is the quantile function calculated from the logit of the probability. It is defined for lp in (-Inf, +Inf) by: $$ qlkiener7(lp, coefk) = m + \frac{\sqrt{3}}{\pi}*g*k*\left(-exp\left(-\frac{lp}{a} +\frac{lp}{w}\right)\right)%qlkiener7(lp, coefk) = m + sqrt(3)/pi*g*k*(-exp(-lp/a) +exp(lp/w)) $$ `varkiener7` designates the Value a-risk and turns negative numbers into positive numbers with the following rule: $$ varkiener7 <- if\;(p <= 0.5)\;\; (- qkiener7)\;\; else\;\; (qkiener7)%varkiener7 <- if (p <= 0.5) (- qkiener7) else (qkiener7) $$ Usual values in finance are `p = 0.01`, `p = 0.05`, `p = 0.95` and `p = 0.99`. `lower.tail = FALSE` uses `1-p` rather than `p`. `ltmkiener7`, `rtmkiener7` and `eskiener7` are respectively the left tail mean, the right tail mean and the expected shortfall of the distribution (sometimes called average VaR, conditional VaR or tail VaR). Left tail mean is the integrale from `-Inf` to `p` of the quantile function `qkiener7` divided by `p`. Right tail mean is the integrale from `p` to `+Inf` of the quantile function `qkiener7` divided by 1-p. Expected shortfall turns negative numbers into positive numbers with the following rule: $$ eskiener7 <- if\;(p <= 0.5)\;\; (- ltmkiener7)\;\; else\;\; (rtmkiener7)%eskiener7 <- if(p <= 0.5) (- ltmkiener7) else (rtmkiener7) $$ Usual values in finance are `p = 0.01`, `p = 0.025`, `p = 0.975` and `p = 0.99`. `lower.tail = FALSE` uses `1-p` rather than `p`. `dtmqkiener7` is the difference between the left tail mean and the quantile when (p <= 0.5) and the difference between the right tail mean and the quantile when (p > 0.5). It is in quantile unit and is an indirect measure of the tail curvature. ## Examples ```r head(ED <- fatreturns(extractData())) (coefk <- paramkienerX(ED, dgts = 3)) x <- -4 xx <- -4:4 p <- 0.1 pp <- pprobs2 dkiener7(x) dkiener7(x, coefk) dkiener7(xx) dkiener7(xx, coefk) pkiener7(x) pkiener7(x, coefk) pkiener7(xx) pkiener7(xx, coefk) qkiener7(p) qkiener7(p, coefk) qkiener7(pp) qkiener7(pp, coefk) rkiener7(10) rkiener7(10, coefk) varkiener7(p) varkiener7(p, coefk) varkiener7(pp) varkiener7(pp, coefk) ltmkiener7(p) ltmkiener7(p, coefk) ltmkiener7(pp) ltmkiener7(pp, coefk) eskiener7(p) eskiener7(p, coefk) eskiener7(pp) eskiener7(pp, coefk) ``` ## References P. Kiener, Explicit models for bilateral fat-tailed distributions and applications in finance with the package FatTailsR, 8th R/Rmetrics Workshop and Summer School, Paris, 27 June 2014. Download it from: [https://www.inmodelia.com/exemples/2014-0627-Rmetrics-Kiener-en.pdf](https://www.inmodelia.com/exemples/2014-0627-Rmetrics-Kiener-en.pdf) P. Kiener, Fat tail analysis and package FatTailsR, 9th R/Rmetrics Workshop and Summer School, Zurich, 27 June 2015. Download it from: [https://www.inmodelia.com/exemples/2015-0627-Rmetrics-Kiener-en.pdf](https://www.inmodelia.com/exemples/2015-0627-Rmetrics-Kiener-en.pdf) C. Acerbi, D. Tasche, Expected shortfall: a natural coherent alternative to Value at Risk, 9 May 2001. Download it from: [https://www.bis.org/bcbs/ca/acertasc.pdf](https://www.bis.org/bcbs/ca/acertasc.pdf) ## See Also Symmetric Kiener distribution K1 `kiener1`, asymmetric Kiener distributions K2, K3 and K4 `kiener2`, `kiener3`, `kiener4`, conversion functions `aw2k`, estimation function `paramkienerX`, estimation function `fitkienerX`, regression function `regkienerLX`.

kiener7 function