分数填补下线性模型响应变量差异的经验似然置信区间

Empirical Likelihood Confidence Intervals for Various Differences of Response Variables in Linear Models Based on Fractional Imputation

在一定条件下证明了缺失数据情形基于分数填补方法得到的线性模型响应变量一般差异指标的经验似然比统计量的渐近分布为加权χ2(1),由此可构造差异指标的经验似然置信区间.

Under certain mild conditions on fractional imputation, it is shown that the asymptotic distribution of the empirical likelihood ratio for various differences of response variables in linear models is scalar χ^2（1）,which can be used to construct empirical likelihood confidence intervals for the differences.