Multiple Imputation After 18+ Years

By Donald B. Rubin. 1996

Abstract Multiple imputation was designed to handle the problem of missing data in public-use data bases where the data-base constructor and the ultimate user are distinct entities. The objective is valid frequency inference for ultimate users who in general have access only to complete-data software and possess limited knowledge of specific reasons and models for nonresponse. For this situation and objective, I believe that multiple imputation by the data-base constructor is the method of choice. This article first provides a description of the assumed context and objectives, and second, reviews the multiple imputation framework and its standard results. These preliminary discussions are especially important because some recent commentaries on multiple imputation have reflected either misunderstandings of the practical objectives of multiple imputation or misunderstandings of fundamental theoretical results. Then, criticisms of multiple imputation are considered, and, finally, comparisons are made to alt...

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Informing sequential clinical decision-making through reinforcement learning: an empirical study (2010, methodology): “4.1 below, we apply multiple imputation (Little and Rubin 1987; Rubin 1996; Schafer 1999) to use all the available data (individuals with both complete and incomplete data) in order to reduce bias.”
Multiple imputation: review of theory, implementation and software (2007, methodology): “New development of MI methodology has proliferated in the literature [1–13].”
Intervention Research in Social Work: Recent Advances and Continuing Challenges (2004, background): “Often we wish to impute a value for missing information so that it can be included in the analysis (Rubin, 1996).”
How to handle missing data: A comparison of different approaches (2015, unknown): “The last two methods are MI techniques and have been identiﬁed as adequate methods to handle missing data (Rubin, 1996; Sterne et al., 2009).”
Women Coming Home: Long‐Term Patterns of Recidivism (2010, unknown): “The process of multiple imputation uses known values on other variables to impute values for those missing cases (Rubin, 1996).”
Computational Strategies for Multivariate Linear Mixed-Effects Models With Missing Values (2002, methodology): “The performance of multiple imputation when the imputer’s and analyst’s models differ was addressed by Meng (1994) and Rubin (1996).”
The effects of parental reading socialization and early school involvement on children's academic performance: a panel study of primary school pupils in the Netherlands (2011, methodology): “To deal with missing values, we employ a multiple imputation procedure in STATA (Rubin, 1987, 1996; Allison, 2000).”
Strategic Eurosceptics and polite xenophobes: Support for the United Kingdom Independence Party (UKIP) in the 2009 European Parliament elections (2012, methodology): “To reduce the level of missing data we imputed the missing values on each attitudinal factor using a multiple imputation procedure (Little & Rubin 1987; Rubin 1996).”
World income inequality databases: an assessment of WIID and SWIID (2015, unknown): “I fit two sets of OLS regressions, one taking account of imputation variability when calculating standard errors (using Rubin’s Rules discussed earlier) and the other ignoring it.”
Using simulation studies to evaluate statistical methods (2017, background): “Of course, randomisation validity would usually be preferred over confidence validity because it implies shorter intervals – but this is not always the case! There are examples of procedures that return both shorter intervals and higher coverage (see for example (45, 46)).”
Universals and specifics in learning strategies: Explaining adolescent mathematics, science, and reading achievement across 34 countries ☆ (2007, methodology): “imputation, we addressed these problems more effectively than other approaches (such as deletion, mean substitution, simple imputation, Peugh & Enders, 2004; Rubin, 1996 ).”
The Relationship Between Citizen Perceptions of Collective Efficacy and Neighborhood Violent Crime (2015, methodology): “Using data from the Project on Human Development in Chicago Neighborhoods (PHDCN), Sampson et al. (1997) validated the proposed relationship between collective efficacy, community structural characteristics, and violence. The work of Sampson et al. (1997) and the work of others using the PHDCN data ”
A weakly informative default prior distribution for logistic and other regression models (2008, methodology): “Multiple imputation [Rubin (1978, 1996)] is another context in which regressions with many predictors are fit in an automatic way.”