SHASH: The Sinh-Arcsinh (SHASH) distribution for fitting a GAMLSS

In Stan125/gamlss.dist: Distributions for Generalized Additive Models for Location Scale and Shape

Description

The Sinh-Arcsinh (SHASH) distribution is a four parameter distribution, for a gamlss.family object to be used for a GAMLSS fitting using the function gamlss(). The functions dSHASH, pSHASH, qSHASH and rSHASH define the density, distribution function, quantile function and random generation for the Sinh-Arcsinh (SHASH) distribution.

There are 3 different SHASH distributions implemented in GAMLSS.

Usage

SHASH(mu.link = "identity", sigma.link = "log", nu.link = "log", 
      tau.link = "log")
dSHASH(x, mu = 0, sigma = 1, nu = 0.5, tau = 0.5, log = FALSE)
pSHASH(q, mu = 0, sigma = 1, nu = 0.5, tau = 0.5, lower.tail = TRUE, 
     log.p = FALSE)
qSHASH(p, mu = 0, sigma = 1, nu = 0.5, tau = 0.5, lower.tail = TRUE, 
     log.p = FALSE)
rSHASH(n, mu = 0, sigma = 1, nu = 0.5, tau = 0.5)

SHASHo(mu.link = "identity", sigma.link = "log", nu.link = "identity", 
      tau.link = "log")
dSHASHo(x, mu = 0, sigma = 1, nu = 0, tau = 1, log = FALSE)
pSHASHo(q, mu = 0, sigma = 1, nu = 0, tau = 1, lower.tail = TRUE, 
     log.p = FALSE)
qSHASHo(p, mu = 0, sigma = 1, nu = 0, tau = 1, lower.tail = TRUE, 
     log.p = FALSE)
rSHASHo(n, mu = 0, sigma = 1, nu = 0, tau = 1)

SHASHo2(mu.link = "identity", sigma.link = "log", nu.link = "identity", 
      tau.link = "log")
dSHASHo2(x, mu = 0, sigma = 1, nu = 0, tau = 1, log = FALSE)
pSHASHo2(q, mu = 0, sigma = 1, nu = 0, tau = 1, lower.tail = TRUE, 
     log.p = FALSE)
qSHASHo2(p, mu = 0, sigma = 1, nu = 0, tau = 1, lower.tail = TRUE, 
     log.p = FALSE)
rSHASHo2(n, mu = 0, sigma = 1, nu = 0, tau = 1)

Arguments

mu.link	Defines the mu.link, with "identity" link as the default for the mu parameter.
sigma.link	Defines the sigma.link, with "log" link as the default for the sigma parameter.
nu.link	Defines the nu.link, with "log" link as the default for the nu parameter.
tau.link	Defines the tau.link, with "log" link as the default for the tau parameter.
x,q	vector of quantiles
mu	vector of location parameter values
sigma	vector of scale parameter values
nu	vector of skewness nu parameter values
tau	vector of kurtosis tau parameter values
log, log.p	logical; if TRUE, probabilities p are given as log(p).
lower.tail	logical; if TRUE (default), probabilities are P[X <= x], otherwise, P[X > x]
p	vector of probabilities.
n	number of observations. If length(n) > 1, the length is taken to be the number required

Details

The probability density function of the Sinh-Arcsinh distribution, (SHASH), Jones(2005), is defined as

f(y|mu,sigma,nu,tau)=c/(sqrt(2*pi)*sigma*(1+z^2)^(1/2)) exp(-(r^2)/2)

where

r=0.5*[exp(tau*sinh^(-1)(z))-exp(-nu*sinh^(-1)(z))]

and

c=0.5*[rho*exp(tau*sinh^(-1)(z))+ nu * exp(-nu*sinh^(-1)(z))]

and z=(y-mu)/sigma for 0<y<0, mu=(-Inf,+Inf), sigma>0, nu>0 and tau>0.

The parameters mu and sigma are the location and scale of the distribution. The parameter nu determines the left hand tail of the distribution with nu>1 indicating a lighter tail than the normal and ν<1 heavier tail than the normal. The parameter tau determines the right hand tail of the distribution in the same way.

The second form of the Sinh-Arcsinh distribution can be found in Jones and Pewsey (2009, p.2) denoted by SHASHo and the probability density function is defined as,

f(y|mu,sigma,nu,tau)= (tau/sigma)* (c/sqrt(2*pi)) 1/(2*(1+z^2)^(1/2)) exp(-(r^2/2))

where

r= sinh(tau*asinh(z)-nu)

and

c= cosh(τ asinh(z)-ν)

and z=(y-mu)/sigma for 0<y<0, mu=(-Inf,+Inf), sigma>0, nu=(-Inf,+Inf) and tau>0.

The third form of the Sinh-Arcsinh distribution (Jones and Pewsey, 2009, p.8) divides the distribution by sigma for the density of the unstandardized variable. This distribution is denoted by SHASHo2 and has pdf

f(y|μ,σ,ν,τ)= (c/sigma)*(tau/((2 π)^(1/2)))*(1/((1+z^2)^(1/2)))-exp(-r^2*0.5)

where z=(y-mu)/(sigma*tau), with r and c as for the pdf of the SHASHo distribution, for 0<y<0, mu=(-Inf,+Inf), sigma>0, nu=(-Inf,+Inf) and tau>0.

Value

SHASH() returns a gamlss.family object which can be used to fit the SHASH distribution in the gamlss() function. dSHASH() gives the density, pSHASH() gives the distribution function, qSHASH() gives the quantile function, and rSHASH() generates random deviates.

Warning

The qSHASH and rSHASH are slow since they are relying on golden section for finding the quantiles

Author(s)

Bob Rigby, Mikis Stasinopoulos [e-mail] and Fiona McElduff

References

Jones, M. C. (2006) p 546-547 in the discussion of Rigby, R. A. and Stasinopoulos D. M. (2005) Appl. Statist., 54, part 3.

Jones and Pewsey (2009) Sinh-arcsinh distributions. Biometrika. 96(4), pp. 761?780.

Rigby, R. A. and Stasinopoulos D. M. (2005). Generalized additive models for location, scale and shape,(with discussion), Appl. Statist., 54, part 3, pp 507-554.

Stasinopoulos D. M. Rigby R. A. and Akantziliotou C. (2006) Instructions on how to use the GAMLSS package in R. Accompanying documentation in the current GAMLSS help files, (see also http://www.gamlss.org/).

Stasinopoulos D. M. Rigby R.A. (2007) Generalized additive models for location scale and shape (GAMLSS) in R. Journal of Statistical Software, Vol. 23, Issue 7, Dec 2007, http://www.jstatsoft.org/v23/i07.

Stasinopoulos D. M., Rigby R.A., Heller G., Voudouris V., and De Bastiani F., (2017) Flexible Regression and Smoothing: Using GAMLSS in R, Chapman and Hall/CRC.

See Also

gamlss.family, JSU, BCT

Examples

SHASH()   # 
plot(function(x)dSHASH(x, mu=0,sigma=1, nu=1, tau=2), -5, 5, 
 main = "The SHASH  density mu=0,sigma=1,nu=1, tau=2")
plot(function(x) pSHASH(x, mu=0,sigma=1,nu=1, tau=2), -5, 5, 
 main = "The BCPE  cdf mu=0, sigma=1, nu=1, tau=2")
dat<-rSHASH(100,mu=10,sigma=1,nu=1,tau=1.5)
hist(dat)
# library(gamlss)
# gamlss(dat~1,family=SHASH, control=gamlss.control(n.cyc=30))

Stan125/gamlss.dist documentation built on May 12, 2019, 7:38 a.m.