摘要
采用贝叶斯方法分析了单指数模型，该方法是通过Reversible Jump Markov Chain MonteCarlo技术（RJMCMC）来实现的．为了获得较快的运算法则，对误差方差和样条系数选取共轭的逆Gamma--正态先验分布，方便地获得其他未知量的边际后验分布并作为目标分布．为了实现从指数向量的条件后验分布中进行抽样的目的，另外设计了一个随机游动（Random Walk）Metropolis抽样器．应用所提议的方法分析了实际数据和例子．
We apply a completed Bayesian method to estimate the single index models and the suggested method is achieved by means of the technique about Reversible Jump Markov Chain Monte Cairo. To obtain the marginal posterior of the unknown elements and to attain a faster algorithm, we specify the conjugate inverse gamma-normal priors for the error variance and spline coefficients. More important for the purpose of quickly sampling from the conditional posterior distribution of the index vectors, we devise a （ random walk ） Metropolis sampler. The real example and simulated example are demonstrated.
出处
《北华大学学报（自然科学版）》
CAS
2010年第2期169-176,共8页
Journal of Beihua University（Natural Science）
