Advanced search
1 file | 553.00 KB Add to list

Mediation analysis in AB/BA crossover studies

Author
Organization
Abstract
Crossover trials are widely used in psychological and medical research to assess the effect of reversible exposures. In such designs, each subject is randomly allocated to a sequence of conditions, enabling the evaluation of treatment differences within each individual. When there are but two possible exposures -each assessed during one of two time periods-, the crossover study is referred to as an AB/BA design. The goal of this presentation is to discuss mediation analysis in such simple crossover studies. We do so by considering within-subject mediation from a counterfactual-based perspective and by deriving expressions for the direct and indirect effects. Employing simulation studies, the performance of several existing methods will be assessed and compared to that of a novel one we propose. We show that the new method yields unbiased and efficient estimators for the direct and indirect effect, under a minimalistic set of `no unmeasured confounding'-assumptions. Finally, we illustrate the different techniques with data from a neurobehavioral study.
Keywords
counterfacuals, AB/BA, Madiation analysis

Downloads

  • M3 May2014.pdf
    • full text
    • |
    • open access
    • |
    • PDF
    • |
    • 553.00 KB

Citation

Please use this url to cite or link to this publication:

MLA
Josephy, Haeike, et al. “Mediation Analysis in AB/BA Crossover Studies.” Modern Modeling Methods Conference, Abstracts, University of Connecticut. Neag School of Education, 2014.
APA
Josephy, H., Loeys, T., & Vansteelandt, S. (2014). Mediation analysis in AB/BA crossover studies. In Modern Modeling Methods conference, Abstracts. Storss, CT, USA: University of Connecticut. Neag School of Education.
Chicago author-date
Josephy, Haeike, Tom Loeys, and Stijn Vansteelandt. 2014. “Mediation Analysis in AB/BA Crossover Studies.” In Modern Modeling Methods Conference, Abstracts. Storss, CT, USA: University of Connecticut. Neag School of Education.
Chicago author-date (all authors)
Josephy, Haeike, Tom Loeys, and Stijn Vansteelandt. 2014. “Mediation Analysis in AB/BA Crossover Studies.” In Modern Modeling Methods Conference, Abstracts. Storss, CT, USA: University of Connecticut. Neag School of Education.
Vancouver
1.
Josephy H, Loeys T, Vansteelandt S. Mediation analysis in AB/BA crossover studies. In: Modern Modeling Methods conference, Abstracts. Storss, CT, USA: University of Connecticut. Neag School of Education; 2014.
IEEE
[1]
H. Josephy, T. Loeys, and S. Vansteelandt, “Mediation analysis in AB/BA crossover studies,” in Modern Modeling Methods conference, Abstracts, Storss, CT, USA, 2014.
@inproceedings{4443344,
  abstract     = {{Crossover trials are widely used in psychological and medical research to assess the effect of reversible exposures. In such designs, each subject is randomly allocated to a sequence of conditions, enabling the evaluation of treatment differences within each individual. When there are but two possible exposures -each assessed during one of two time periods-, the crossover study is referred to as an AB/BA design. The goal of this presentation is to discuss mediation analysis in such simple crossover studies. We do so by considering within-subject mediation from a counterfactual-based perspective and by deriving expressions for the direct and indirect effects. Employing simulation studies, the performance of several existing methods will be assessed and compared to that of a novel one we propose. We show that the new method yields unbiased and efficient estimators for the direct and indirect effect, under a minimalistic set of `no unmeasured confounding'-assumptions. Finally, we illustrate the different techniques with data from a neurobehavioral study.}},
  author       = {{Josephy, Haeike and Loeys, Tom and Vansteelandt, Stijn}},
  booktitle    = {{Modern Modeling Methods conference, Abstracts}},
  keywords     = {{counterfacuals,AB/BA,Madiation analysis}},
  language     = {{eng}},
  location     = {{Storss, CT, USA}},
  publisher    = {{University of Connecticut. Neag School of Education}},
  title        = {{Mediation analysis in AB/BA crossover studies}},
  year         = {{2014}},
}