Multiple Imputation of Predictor Variables Using Generalized Additive Models
Multiple Imputation of Predictor Variables Using Generalized Additive Models
复制标题
DOI：
10.1080/03610918.2014.911894
复制
发表时间：
2016-01-01
影响因子：
0.9
通讯作者：
Spiess, Martin
中科院分区：
文献类型：
--
作者：
De Jong, Roel;Van Buuren, Stef;Spiess, Martin
通过模拟研究了多种imputation方法对偏离分布假设的敏感性，其中科学兴趣的参数是线性回归模型的系数，预测变量的值是随机缺失的。研究了一种基于位置、尺度和形状广义加性模型（GAMLSS）的新方法的性能。尽管基于预测均值匹配的imputation方法实际上是无偏的，但即使在所有变量都是联合正态分布的实验中，它们也存在轻度到中度的覆盖不足。GAMLSS方法比目前可用的方法具有更好的覆盖范围。
The sensitivity of multiple imputation methods to deviations from their distributional assumptions is investigated using simulations, where the parameters of scientific interest are the coefficients of a linear regression model, and values in predictor variables are missing at random. The performance of a newly proposed imputation method based on generalized additive models for location, scale, and shape (GAMLSS) is investigated. Although imputation methods based on predictive mean matching are virtually unbiased, they suffer from mild to moderate under-coverage, even in the experiment where all variables are jointly normal distributed. The GAMLSS method features better coverage than currently available methods.