基于偏正态数据下联合位置与尺度混合专家回归模型的参数估计

The parameter estimation of the mixture of expert regression models for joint location and scale with skew-normal data

混合专家回归模型广泛应用于异质总体数据的分类,聚类及回归分析中.研究基于偏正态数据,提出了联合位置与尺度混合专家回归模型,该模型同时对位置,尺度和混合比例参数建模,应用MM算法和EM算法研究了该模型参数的极大似然估计.通过随机模拟和实例分析说明了该模型和方法的有效性与实用性.

The mixture of expert regression models are widely used for classification of hetero-geneous data,clustering and regression analysis.Based on the skew-normal data,the mixture expert regression models for joint location and scale are proposed,which is used to simultaneously model the location,scale and mixture proportion parameters of interest.MM algorithm and EM algorithm are considered to estimate unknown parameters in the maximum likelihood estimation procedure.Monte Carlo simulation and a real data are used to verify the proposed method.