摘要
纵向研究广泛应用于医学、生物学等研究领域。纵向数据为典型的相关数据,数据方差结构选择的正确与否直接影响估计的效率。当样本量较大时,QIF方法表明工作相关阵的选择等价于无偏估计方程的选择。本文主要通过模拟讨论样本量较小时,利用经验相对效率选择无偏估计方程。模拟表明,如果增加没有信息的无偏估计方程,估计的效率就会减少,并且随着样本量的增加,估计的效率也会得到提高。
Longitudinal study is widely applied into medical and biological field.Longitudinal data is typical related data,and it is desirable to choose correct data variance structure since the correct structure affects the efficiency of estimation.When the sample size is great,QIF approach shows that the selection of working correlation is same to the selection of unbiased estimating equation.When the sample size is small,we choose the unbiased estimating equation using empirical relative efficiencies.Simulation shows that the increasing of unbiased estimating equation without information would reduce the efficiency of the estimation,and the efficiency would improve when the sample size is increased.
