Bayesian estimation of \(g\)-and-\(k\) distributions using MCMC (Q2488375)

The generalized \(g\)-and-\(k\) distribution can be characterized by it's quantile function \[ Q(u)=A+B\left( 1+c{1-e^{-gz_u}\over 1+e^{-gz_u}} \right)z_u(1+z_u^2)^k, \] where \(A\) and \(B\) are location and scale parameters, \(g\in R\) is a measure of skewness, \(k>-0.5\) is a measure of kurtosis, \(c=0.8\), and \(z_u\) is the \(u\)-th quantile of the standard normal distribution. The authors describe a Bayes technique for estimation of \((A,B,g,k)\). Markov chain Monte Carlo with a Metropolis-Hastings sampler is used for the posterior evaluation. Results of simulations and application to medical data are presented.