Parameter estimation of α-stable distributions using DRAM algorithm

郝燕玲,单志明,沈锋

DOI: https://doi.org/10.13245/j.hust.2011.10.012

2011-01-01

Abstract:Jointly estimation of α-stable density parameters was aimed,and a method was proposed based on Bayesian inference and Markov chain Monte Carlo（MCMC） approach,by combining the adaptive Metropolis（AM） sampling and delayed rejection（DR） algorithm.All unknown parameters were viewed as variables under Bayesian framework in order to change the parameter estimation into probability calculation.The global and local adaptations combined strategy could automatically and quickly construct an approximation of the target distribution,which increased the efficiency of the sampler for MCMC approach.Simulation results indicate that the proposed method can estimate the four parameters of α-stable distributions and performs more accurately and robustly.