Sensitivity analysis after multiple imputation under missing at random: a weighting approach

Carpenter, James R.; Kenward, Michael G.; White, Ian R.
June 2007
Statistical Methods in Medical Research;Jun2007, Vol. 16 Issue 3, p259
Academic Journal
journal article
Multiple imputation (MI) is now well established as a flexible, general, method for the analysis of data sets with missing values. Most implementations assume the missing data are ;missing at random' (MAR), that is, given the observed data, the reason for the missing data does not depend on the unseen data. However, although this is a helpful and simplifying working assumption, it is unlikely to be true in practice. Assessing the sensitivity of the analysis to the MAR assumption is therefore important. However, there is very limited MI software for this. Further, analysis of a data set with missing values that are not missing at random (NMAR) is complicated by the need to extend the MAR imputation model to include a model for the reason for dropout. Here, we propose a simple alternative. We first impute under MAR and obtain parameter estimates for each imputed data set. The overall NMAR parameter estimate is a weighted average of these parameter estimates, where the weights depend on the assumed degree of departure from MAR. In some settings, this approach gives results that closely agree with joint modelling as the number of imputations increases. In others, it provides ball-park estimates of the results of full NMAR modelling, indicating the extent to which it is necessary and providing a check on its results. We illustrate our approach with a small simulation study, and the analysis of data from a trial of interventions to improve the quality of peer review.


Related Articles

  • Evaluation of software for multiple imputation of semi-continuous data. Yu, L.-M.; Burton, Andrea; Rivero-Arias, Oliver // Statistical Methods in Medical Research;Jun2007, Vol. 16 Issue 3, p243 

    It is now widely accepted that multiple imputation (MI) methods properly handle the uncertainty of missing data over single imputation methods. Several standard statistical software packages, such as SAS, R and STATA, have standard procedures or user-written programs to perform MI. The...

  • Review of inverse probability weighting for dealing with missing data. Seaman, Shaun R; White, Ian R // Statistical Methods in Medical Research;Jun2013, Vol. 22 Issue 3, p278 

    The simplest approach to dealing with missing data is to restrict the analysis to complete cases, i.e. individuals with no missing values. This can induce bias, however. Inverse probability weighting (IPW) is a commonly used method to correct this bias. It is also used to adjust for unequal...

  • Alternative analyses for handling incomplete follow-up in the intention-to-treat analysis: the randomized controlled trial of balloon kyphoplasty versus non-surgical care for vertebral compression fracture (FREE).  // BMC Medical Research Methodology;2012, Vol. 12 Issue 1, p35 

    The article presents a randomized controlled trial of balloon kyphoplasty (BKP) versus non-surgical care for vertebral compression fracture (FREE). As stated, during the study percutaneous BKP was performed with introducer tools, inflatable bone tamps, and polymethylmethacrylate bone cement on...

  • Missing covariate data in clinical research: when and when not to use the missing-indicator method for analysis. Groenwold, Rolf H. H.; White, Ian R.; Donders, A. Rogier T.; Carpenter, James R.; Altman, Douglas G.; Moons, Karel G. M. // CMAJ: Canadian Medical Association Journal;8/7/2012, Vol. 184 Issue 11, p1265 

    The article presents a review of the missing-indicator method and illustrate its validity with the use of real data with incomplete covariates from randomized and nonrandomized studies. It states that the easiest way to handle missing covariate data is to omit from participants with any missing...

  • Missing Data. Newgard, Craig D.; Lewis, Roger J. // JAMA: Journal of the American Medical Association;9/1/2015, Vol. 314 Issue 9, p940 

    The article looks at the possible imputation methodologies for handling missing data in a clinical study so as to determine the most valid results and to avoid introducing bias, focusing on an article by George L. Bakris and colleagues in the current issue reporting their randomized clinical...

  • Missing Data in Randomized Clinical Trials for Weight Loss: Scope of the Problem, State of the Field, and Performance of Statistical Methods. Elobeid, Mai A.; Padilla, Miguel A.; McVie, Theresa; Thomas, Olivia; Brock, David W.; Musser, Bret; Kaifeng Lu; Coffey, Christopher S.; Desmond, Renee A.; St-Onge, Marie-Pierre; Gadde, Kishore M.; Heymsfield, Steven B.; Allison, David B. // PLoS ONE;2009, Vol. 4 Issue 8, p1 

    Background: Dropouts and missing data are nearly-ubiquitous in obesity randomized controlled trails, threatening validity and generalizability of conclusions. Herein, we meta-analytically evaluate the extent of missing data, the frequency with which various analytic methods are employed to...

  • OUTLIER ROBUST POSTERIOR PREDICTIVE CHECKS FOR MISSING DATA MODELS. NICORICI, Galina // Economic Computation & Economic Cybernetics Studies & Research;2014, Vol. 48 Issue 1, p1 

    The goal of this paper is to adjust the posterior predictive checks in order to overcome the problem of simultaneous outliers and missing data. We combine the approaches used to treat these two problems (separately) in the construction of particular posterior predictive p-values for Bayesian...

  • MISSING DATA IMPUTATION IN CARDIAC DATA SET (SURVIVAL PROGNOSIS). Kumar, R.Kavitha; Chadrasekar, R. M. // International Journal on Computer Science & Engineering;2010, Vol. 2 Issue 5, p1836 

    Treating missing value is very big task in the data preprocessing methods. Missing data are a potential source of bias when analyzing clinical trials. In this paper we analyze the performance of different data imputation methods in a task where the aim is to predict the probability of survival...

  • Multiple imputation to deal with missing EQ-5D-3L data: Should we impute individual domains or the actual index? Simons, Claire; Rivero-Arias, Oliver; Yu, Ly-Mee; Simon, Judit // Quality of Life Research;Apr2015, Vol. 24 Issue 4, p805 

    Purpose: Missing data are a well-known and widely documented problem in cost-effectiveness analyses alongside clinical trials using individual patient-level data. Current methodological research recommends multiple imputation (MI) to deal with missing health outcome data, but there is little...


Read the Article


Sorry, but this item is not currently available from your library.

Try another library?
Sign out of this library

Other Topics