Patchwork rejection algorithms (Q915340)

Let U, R be independent uniform random variables with support [0,1], let f(x) be the density function of the required target distribution and let g(x) be the density of an envelope distribution which belongs to a family of continuous distribution functions G(x). Then \(x=G^{-1}(u)\) generates a random envelope variate. The condition for the acceptance of x as a random target variate is given by \(r\cdot c\cdot g(G^{-1}(u))\leq f(G^{-1}(u))\) where \(c:=\sup \{f(G^{-1}(u))/g(G^{-1}(u)):\) \(u\}<\infty\). Some examples are discussed in the case of a normal distribution for f and for the family of Lomax distributions or Rayleigh distributions or exponential distributions.