A Generator for Generalized Inverse Gaussian Distributions
Xiaozhu Zhang, Jerome P. Reiter
Abstract
Open-access reader
Xiaozhu Zhang, Jerome P. Reiter
Abstract
Open-access reader
We propose a new generator for the generalized inverse Gaussian (GIG) distribution by decomposing the density of GIG into two components. The first component is a truncated inverse Gamma density, in order to sample from which we improve the traditional inverse CDF method. The second component is the product of an exponential pdf and an inverse Gamma CDF. In order to sample from this quasi-density, we develop a rejection sampling procedure that adaptively adjusts the piecewise proposal density according to the user-specified rejection rate or the desired number of cutoff points. The resulting complete algorithm enjoys controllable rejection rate and moderate setup time. It preserves efficiency for both parameter varying case and large sample case.
OpenAlex reports 1 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
We propose a new generator for the generalized inverse Gaussian (GIG) distribution by decomposing the density of GIG into two components. The first component is a truncated inverse Gamma density, in order to sample from which we improve the traditional inverse CDF method. The second component is the product of an exponential pdf and an inverse Gamma CDF. In order to sample from this quasi-density, we develop a rejection sampling procedure that adaptively adjusts the piecewise proposal density according to the user-specified rejection rate or the desired number of cutoff points. The resulting complete algorithm enjoys controllable rejection rate and moderate setup time. It preserves efficiency for both parameter varying case and large sample case.
Key concepts: Inverse Gaussian distribution, Generalized inverse Gaussian distribution, Inverse, Inverse-gamma distribution, Generator (circuit theory), Piecewise, Gaussian, Mathematics