IJE Advance Access originally published online on August 27, 2004
International Journal of Epidemiology 2005 34(1):89-99; doi:10.1093/ije/dyh297
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
IJE vol.34 no.1 © International Epidemiological Association 2004; all rights reserved.
Article |
Comparison of imputation and modelling methods in the analysis of a physical activity trial with missing outcomes
1 MRC Biostatistics Unit, Institute of Public Health, Robinson Way, Cambridge CB2 2SR, UK
2 Department of Epidemiology and Public Health, University College London, 1–19 Torrington Place, London WC1E 6BT, UK
3 Medical Statistics Unit, London School of Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT, UK
Correspondence: Dr Angela Wood, MRC Biostatistics Unit, Institute of Public Health, Robinson Way, Cambridge CB2 2SR, UK. E-mail: angela.wood{at}mrc-bsu.cam.ac.uk
Background Longitudinal studies almost always have some individuals with missing outcomes. Inappropriate handling of the missing data in the analysis can result in misleading conclusions. Here we review a wide range of methods to handle missing outcomes in single and repeated measures data and discuss which methods are most appropriate.
Methods Using data from a randomized controlled trial to compare two interventions for increasing physical activity, we compare complete-case analysis; ad hoc imputation techniques such as last observation carried forward and worst-case; model-based imputation; longitudinal models with random effects; and recently proposed joint models for repeated measures data and non-ignorable dropout.
Results Estimated intervention effects from ad hoc imputation methods vary widely. Standard multiple imputation and longitudinal modelling agree closely, as they should. Modifying the modelling method to allow for non-ignorable dropout had little effect on estimated intervention effects, but imputing using a common imputation model in both groups gave more conservative results.
Conclusions Results from ad hoc imputation methods should be avoided in favour of methods with more plausible assumptions although they may be computationally more complex. Although standard multiple imputation methods and longitudinal modelling methods are equivalent for estimating the treatment effect, the two approaches suggest different ways of relaxing the assumptions, and the choice between them depends on contextual knowledge.
Keywords Missing data, dropout, last observation carried forward, imputation, longitudinal data, random effects, non-ignorable
Accepted 30 June 2004
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  M. A. Klebanoff and S. R. Cole Use of Multiple Imputation in the Epidemiologic Literature Am. J. Epidemiol., August 15, 2008; 168(4): 355 - 357. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. von Huth Smith, S. Ladelund, K. Borch-Johnsen, and T. Jorgensen A randomized multifactorial intervention study for prevention of ischaemic heart disease (Inter99): The long-term effect on physical activity Scand J Public Health, June 1, 2008; 36(4): 380 - 388. [Abstract] [PDF]
|
|
|
|
 |
|
 S. van Buuren Multiple imputation of discrete and continuous data by fully conditional specification Statistical Methods in Medical Research, June 1, 2007; 16(3): 219 - 242. [Abstract] [PDF]
|
|
|
|
 |
|
 I. R White, J. Carpenter, S. Evans, and S. Schroter Eliciting and using expert opinions about dropout bias in randomized controlled trials Clinical Trials, April 1, 2007; 4(2): 125 - 139. [Abstract] [PDF]
|
|
|
|
 |
|
 L. U. Rosenberg, K. Einarsdottir, E. I. Friman, S. Wedren, P. W. Dickman, P. Hall, and C. Magnusson Risk Factors for Hormone Receptor-Defined Breast Cancer in Postmenopausal Women Cancer Epidemiol. Biomarkers Prev., December 1, 2006; 15(12): 2482 - 2488. [Abstract] [Full Text] [PDF]
|
|