What Improves with Increased Missing Data Imputations?

By Todd E Bodner. 2008

665 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/03ce70f9334d91f14e9f26491968add09e7dd2d5

How researchers see this paper

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

In Statistics and Probability

Citations in the corpus

New Confidence Intervals and Bias Comparisons Show That Maximum Likelihood Can Beat Multiple Imputation in Small Samples (2013, neutral): “If the fraction of missing information is large, standard errors estimated by MI can require 200-300 imputed datasets to approach the efficiency of standard errors estimated by ML (Bodner, 2008; Graham et al., 2007).”
Multiple imputation using chained equations for missing data in TIMSS: a case study (2013, neutral): “Increasing the number of imputations has thus led to more precision in computing the p-values, standard errors, confidence intervals, and fractions of missing information (Bodner, 2008).”
The proportion of missing data should not be used to guide decisions on multiple imputation (2019, background): “FMI is a highly variable estimate at low numbers of imputations [30], hence the need for a large number of imputations.”
Pooling ANOVA Results From Multiply Imputed Datasets (2016, methodology): “Following recent recommendations, we created M = 100 imputed datasets for each simulated dataset (Bodner, 2008; Graham, Olchowski, & Gilreath, 2007).”
Missing data and multiple imputation in clinical epidemiological research (2017, neutral): “Later, Bodner 28 and White et al 23 suggested the rule of thumb in order to increase a level of reproducibility of the results in practice; the number of imputations should be similar to the percentage of incomplete cases.”
Missing Data Imputation versus Full Information Maximum Likelihood with Second-Level Dependencies (2011, methodology): “The choice to use a large number of imputations is based on work by Bodner (2008) that concludes that the earlier suggestions for a low number of imputations are actually inappropriate with a large percentage of missing data.”
Multiple Imputation Strategies for Multiple Group Structural Equation Models (2011, background): “, Bodner, 2008; Graham, Olchowski, & Gilreath, 2007). The rationale for using a larger number of imputations is that it improves power. For example, Graham et al. (2007) suggested that analyzing m D 20 imputations generally yields power that is comparable to a full information maximum likelihood ana”
How Many Imputations Do You Need? A Two-stage Calculation Using a Quadratic Rule (2016, background): “A recently proposed rule is that, to estimate SEs reliably, you should choose the number of imputations according to the linear rule 100(cid:2011) (Bodner, 2008; White et al., 2011), where is the fraction of missing information, defined below in equation (4).”
On Obtaining Estimates of the Fraction of Missing Information From Full Information Maximum Likelihood (2012, methodology): “Thus, the method presented here directly addresses the concerns raised by Bodner (2008) with regard to the quality of MI-based estimates of fraction of missing information based on small M , and obviates the need for confidence intervals to evaluate the impact of small M .”
HOW TO IMPUTE INTERACTIONS, SQUARES, AND OTHER TRANSFORMED VARIABLES (2009, methodology): “We applied all of the imputation methods that we described earlier, using M = 40 imputations (Bodner 2008).”
Multiple Imputation of Missing Data for Multilevel Models (2018, background): “…generating 20 imputations is sufficient for most applications in which the primary goal is to estimate the model parameters, but as many as 100 or more imputations can be useful if the analyses involve testing more elaborate statistical hypotheses (Bodner, 2008; see also Grund et al., 2016b).”
Using plausible values in secondary analysis in large-scale assessments (2017, methodology): “Bodner (2008) gave similar recommendations for the estimation of the null hypothesis significance test p-values and confidence interval half-widths.”