Conditional Sampling from a Gamma Distribution given Sufficient Statistics
Abstract
This thesis is an analysis of conditional sampling from a gamma distributiongiven sufficient statistics. Several sampling algorithms are considered. Analgorithm similar to direct sampling is discussed in particular. This algorithmuses parameter adjustments to meet conditions of sufficient statistics. However,this algorithm is influenced by a pivotal condition. How this condition affectsalgorithm 1 is presented. A Gibbs sampler is assumed to give correct samples,and will be used in comparison to the other samplers. Several data sets areused, and all of them follow the case of 3 data points.