Diagnosing and responding to violations in the positivity assumption

Statistical Methods in Medical Research - Tập 21 Số 1 - Trang 31-54 - 2012
Maya Petersen1, Kristin E. Porter1, Susan Gruber1, Yue Wang2, Mark J. van der Laan1
1Division of Biostatistics, University of California, Berkeley, CA 94110-7358, USA.
2Department of Clinical Information Services, Novartis Pharmaceuticals Corporation, East Hanover, New Jersey, USA.

Tóm tắt

The assumption of positivity or experimental treatment assignment requires that observed treatment levels vary within confounder strata. This article discusses the positivity assumption in the context of assessing model and parameter-specific identifiability of causal effects. Positivity violations occur when certain subgroups in a sample rarely or never receive some treatments of interest. The resulting sparsity in the data may increase bias with or without an increase in variance and can threaten valid inference. The parametric bootstrap is presented as a tool to assess the severity of such threats and its utility as a diagnostic is explored using simulated and real data. Several approaches for improving the identifiability of parameters in the presence of positivity violations are reviewed. Potential responses to data sparsity include restriction of the covariate adjustment set, use of an alternative projection function to define the target parameter within a marginal structural working model, restriction of the sample, and modification of the target intervention. All of these approaches can be understood as trading off proximity to the initial target of inference for identifiability; we advocate approaching this tradeoff systematically.

Từ khóa


Tài liệu tham khảo

10.2307/2527916

10.1016/0270-0255(86)90088-6

10.1016/0270-0255(86)90088-6

Robins JM., 1999, In: Proceedings of the American Statistical Association: Section on Bayesian Statistical Science, 6

Wang Y, 2006, Technical Report 211, Division of Biostatistics

Pearl J., 2000, Causality: models, reasoning, and inference

10.1214/ss/1177012031

10.1037/h0037350

Robins JM., 1998, In: Proceedings of the American Statistical Association. Section on Bayesian Statistical Science 1997, 1

Robins JM., 1999, In: Halloran E and Berry D (eds) Statistical models in epidemiology, the environment, and clinical trials (Minneapolis, MN, 1997), 95

10.1016/j.jspi.2005.12.008

10.1016/S0021-9681(87)80018-8

10.1016/j.jspi.2004.06.060

Bembom O, 2007, EJS, 1, 574

Moore KL, 2009, Technical Report 255, Division of Biostatistics

10.1093/aje/kwn164

Rosenblum MM, 2001, Int J Biostat, 1, 4

van der Laan MJ, 2003, Technical Report 130, Division of Biostatistics

van der Laan MJ, 2007, Genet Mol Biol, 6

10.1007/978-0-387-84858-7

10.1097/00001648-200009000-00011

Kish L., 1992, J Official Stat, 8, 183

Bembom O, 2008, Technical Report 230,Division of Biostatstics

Robins JM, 2001, Stat Sinica, 11, 920

10.1002/sim.1110

Scharfstein DO, 1999, J Am Stat Assoc, 94, 1121

10.2202/1557-4679.1043

10.2202/1557-4679.1238

10.1177/0193841X08317586

Petersen ML, 2010, Technical report, division of Biostatstics

10.1007/978-1-4757-2545-2

10.1002/sim.3414

10.2202/1557-4679.1182

Johnson VA, 2009, Top HIV Med, 17, 138

Bembom O, 2008, Technical Report 231, Division of Biostatstics

10.3386/t0330

LaLonde RJ., 1986, Am Econ Rev, 76, 604

10.2307/2971733

10.1080/01621459.1999.10473858

10.2202/1557-4679.1022

10.1002/sim.3301