Testing for Covariate Balance in Comparative Studies.

Covariate Balance;Observational Study;Propensity Score Matching;Statistics and Numeric Data;Science;Statistics

Kleyman, Yevgeniya N.Xie, Yu ;

University of Michigan

关键词: Covariate Balance;    Observational Study;    Propensity Score Matching;    Statistics and Numeric Data;    Science;    Statistics;   

Others  :  https://deepblue.lib.umich.edu/bitstream/handle/2027.42/64769/ykleyman_1.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

In comparative studies, causal inference necessitates effective adjustments forimportant covariates. This becomes particularly relevant in observational studies,where covariates are rarely jointly balanced between treatment and control groups,and this lack of covariate balance can generate misleading results. Such adjustmentsas propensity score matching and stratification are frequently used to align dissimilargroups. The credibility of the analysis may be bolstered by demonstrating that suchan adjustment has improved balance on observed covariates. It is not automaticthat these measures achieve this objective. To appraise whether they have, practitionersuse a variety of techniques, many of them common in hypothesis testing.However, some of these ;;balance tests;; lack formal motivation, may give contradictorymessages about balance, and have in some cases been shown to have undesirablestatistical properties.We begin by identifying goals of balance testing in comparative studies and evaluatingarguments used in the literature to support and oppose using significance tests to appraise covariate balance. We survey the literature for existing appraisals of balance, with an interest in their advantages and shortcomingsin relation to the goals we identify. We study the performance of some existingways to assess covariate balance through an examination of contemporary researchand identify that a permutation version of the balance-testing procedure originallysuggested in Dehejia and Wahba (2002) can outperform some of theother approaches. We supplement our findings from the literature with athorough simulation study based on real observational data. We use this simulationstudy to evaluate the impact of using the various balance diagnostics on the validityof statistical inferences about treatment effects.Our literature review and simulation results suggest an important role for a newformal balance diagnostic. We develop several ideas aimed at this end and test themin various simulation settings that resemble authentic analysis conditions. Using theresults of the literature survey and our simulation study, we are able to recommendnew and existing techniques for testing covariate balance using randomization-basedinference. We also propose dependable combinations of procedures for inference incomparative studies and provide some examples for application of our recommendedtechniques.