零膨胀几何分布的参数估计

Parameter Estimation of Zero-Inflated Geometric Distribution

在产品质量检验过程中,经常会出现零观测值较多的情况.为更好拟合这类数据,提出零膨胀几何分布模型,引入隐变量,运用极大似然估计(MLE)、最大期望(EM)算法下的极大似然估计及贝叶斯估计对模型参数进行估计.设定不同的样本量,不同的参数真值,采用数值模拟方法对上述研究方法的性能进行评估.

There are often more zero observations in the fields of product quality inspection.In order to better fit such data,a zero-inflated geometric distribution model was proposed.Implicit variables were introduced to estimate the model parameters by using maximum likelihood estimation(MLE),maximum likelihood estimation under expectation maximization(EM)algorithm and Bayesian estimation.With different sample sizes and different true values of the parameters,the performances of research methods were evaluated by numerical simulation.