非相容与误设定下多重插补的自举推断
统计方法学
2019-11-28 v2
摘要
多重插补已成为统计分析中处理缺失数据最流行的方法之一。这一成功部分归功于 Rubin 简单的组合规则。当插补与分析过程所谓相容且完整数据分析有效时,这些规则给出频率论有效的推断,否则可能无效。粗略而言,相容性对应于插补模型与分析模型是否对数据作出不同假设。在实践中,插补与分析过程常常不相容,使得检验可能不具有正确的尺度,且置信区间覆盖度偏离宣称的水平。我们考察了若干近期将自举与多重插补结合的提议,并判定哪些在非相容与模型误设定下是有效的。插补后自举通常在非相容或误设定下不能给出有效的方差估计,而自举后插补则可以。我们推荐一种特定的计算高效的自举后插补变体。
引用
@article{arxiv.1911.09980,
title = {Bootstrap Inference for Multiple Imputation under Uncongeniality and Misspecification},
author = {Jonathan W. Bartlett and Rachael A. Hughes},
journal= {arXiv preprint arXiv:1911.09980},
year = {2019}
}
备注
Updated (fixed) reference based simulation results. Now included tables which were previously not included as they were in supplementary information document. Swapped order of the two simulation studies. Added acknowledgement and funding statements