约束函数型数据半参数模型的贝叶斯估计  

Bayesian Estimation for Constrained Partial Functional Linear Regression

作　　者：丁建华 DING Jian-hua(Department of Statistics,KLATASDS,Shanxi Datong University,Datong 037009,China)

机构地区：[1]山西大同大学数学与统计学院,统计与数据科学前沿理论及应用教育部重点实验室,山西大同037009

出　　处：《数学的实践与认识》2021年第18期178-184,共7页Mathematics in Practice and Theory

基　　金：全国统计科学研究项目;教育部重点实验室开放课题。

摘　　要：介绍了函数型数据半参数模型的估计问题,其中斜率函数满足单调性、凸凹性等形状约束条件.通过惩罚样条最小二乘估计推导出线性混合效应模型,进而提出了贝叶斯估计方法,并给出了马尔可夫链蒙特卡洛(MCMC)算法.模拟结果表明所提出的方法是有效的.The paper introduces the estimation method for partial functional linear model when the slope function is subject to a variety of shape constraints such as monotonicity,convexity or concavity,etc.The linear mixed effects model is derived by the penalized splines least square estimate.Bayesian estimate is employed and Markov chain Monte Carlo(MCMC) algorithm is constructed.Simulation results show that the proposed method is effective.

关 键 词：函数型数据 约束惩罚B-样条 MCMC算法 贝叶斯估计 

分 类 号：O212.8[理学—概率论与数理统计]