纵向数据的稳健估计及案例分析

Robust estimation of longitudinal data and case analysis

纵向数据是对一组观测个体按时间或空间顺序重复跟踪监测而得,即对每一个体在不同时间或不同实验条件下进行多次测量,而多次重复观测之间一般具有相关性,会影响对应估计的准确性.研究一般线性模型下纵向数据的理论推导和实例分析,发现模型的最小二乘估计、极大似然估计以及约束极大似然估计的参数估计结果差别不大甚至基本相等,验证了纵向数据估计的稳健性.

Longitudinal data is obtained by repeated tracking and monitoring of a group of observation individuals in time or space sequence,that is,multiple measurements are made for each individual at different times or under different experimental conditions.However,there is a correlation between multiple repeated observations,which will affect the accuracy of corresponding estimation.The theoretical derivation and case analysis of longitudinal data under the general linear model is studied,and it is found that the parameter estimation results of least squares estimation,maximum likelihood estimation and constrained maximum likelihood estimation of the model have little difference or even almost equal,which verifies the robustness of longitudinal data estimation.