“Gibbs抽样中,从第m次到第m+1次的迭代过程并不是简单的从全条件分布中抽取相关参数就可以了,而是需要根据jump distribution 计算所抽取的具体值被接受的概率,然后根据相应的概率确定是否抽取该值。”
请问各位专家是否知道“需要根据jump distribution 计算所抽取的具体值被接受的概率,然后根据相应的概率确定是否抽取该值”是怎么回事?我只是知道在Metropolis Hastings抽样中是以概率决定是否转移
Gibbs Sampler is a special case of M-H with the proposal density given by the full conditionals. But Gibbs always accepts the proposed state as opposed to acceptance/rejection of the M-H.