Imputation of missing values in the case of a multiple item instrument measuring alcohol consumption
Imputation of missing values in the case of a multiple item instrument measuring alcohol consumption
复制标题
DOI：
10.1002/sim.837
复制
发表时间：
2001-08-15
影响因子：
2
通讯作者：
Gmel, G
中科院分区：
文献类型：
--
作者：
Gmel, G
测量仪器的缺失值是困扰调查研究者的一个普遍问题。在用于测量酒精消耗的仪器中，这种情况更为严重：它们通常由项目电池组成，用于构建汇总测量，例如每天的纯酒精克数，并且关于几种饮料中的一种的单一项目的缺失值（例如，数量或频率）导致所有饮料的汇总测量缺失，尽管其他项目的值是已知的。本文探讨了不同的方法来推算缺失值，可行的标准统计软件包。Hot-deck imputation被证明具有一定的优势，但即使是单值imputation（例如，中位数imputation）所得到的值也与其他三种方法的值相当。版权所有John Wiley & Sons， Ltd。
Missing values in survey instruments are a common problem for survey researchers. It is aggravated in the case of instruments used to measure alcohol consumption: they usually consist of item batteries from which summary measures, such as grams of pure alcohol per day, are constructed, and a missing value (for example, quantity or frequency) in regard to a single item for only one of several beverages results in a missing summary measure across all of the beverages, though the values for the remaining items are known. The present paper examines different approaches to imputation of missing values, feasible with standard statistical software packages. Hot-deck imputation is shown to have certain advantages, but even single-value imputation (for example, median imputation) results in values that are comparable to those of the other three imputation methods. Copyright (C) 2001 John Wiley & Sons, Ltd.