Theory and methods for sufficient cause interactions

描述(由申请人提供)：拟议研究的目标是开发测试足够原因相互作用的理论和方法。这些方法将有助于识别生物系统中的机械性相互作用，以及在基因-基因和基因-环境相互作用的遗传流行病学研究的分析和解释中。众所周知，统计模型中交互作用的存在取决于所使用的模型，而且统计交互作用不需要对应于在任何生物或物理意义上有意义的交互作用。充分原因框架提到了产生结果所涉及的实际因果机制，称为充分原因。当两个或两个以上的二元原因参与相同的因果机制时，协同作用就被认为是存在的。有时，无法从数据中确定协同作用；当数据确实意味着必须存在协同作用时，则认为存在充分的原因交互作用。通过拟议的研究开发的理论和方法导致了对充分原因相互作用的经验测试，从而构成了对单一原因机制中两个或更多原因联合存在的测试。这项研究的目的是扩展关于二分暴露的充分原因框架的理论，发展顺序和分类暴露的充分原因交互作用的理论，为存在充分原因交互作用发展多重稳健的半参数检验，并表征对充分原因交互作用检验产生有效结论的暴露错误分类形式。这项研究将提供一套可用于识别生物系统中的机械性相互作用的技术，并将开发一个将这些机械性相互作用概念化的理论框架，并提供对此类相互作用进行经验检验的方法。这项研究将有助于识别基因-基因和基因-环境的相互作用机制，从而增加我们对遗传机制的理解。这项关于充分原因交互作用的研究对于理解标准的基因-基因和基因-环境交互作用测试和研究设计的机制含义的意义将被探索，所开发的方法将被应用于砷的健康影响纵向研究的几个数据集。这项研究将对统计文献中关于相互作用概念和测量误差对因果推断的影响的研究取得重要进展。整个研究计划将有助于我们理解因果关系的概念，这些概念构成了因果推理中统计文献的基础，并被用于医学、流行病学、心理学、遗传学、计算机科学、哲学、社会学、教育和经济学。 公共卫生相关性：拟议研究中开发的统计方法将有助于确定生物系统中的机械性相互作用，并有助于分析和解释基因-基因和基因-环境相互作用的遗传流行病学研究。这些方法将应用于《砷对健康的影响纵向研究》的数据，以提供有关砷暴露可能导致疾病的潜在病理生理学和机制的知识。这项研究将对统计文献中关于相互作用概念和测量误差对因果推断的影响的研究取得重要进展。

DESCRIPTION (provided by applicant): The objective of the proposed research is to develop theory and methods for testing for sufficient cause inter- actions. The methods will be useful in identifying mechanistic interactions in biological systems and in and in the analysis and interpretation of studies in genetic epidemiology of gene-gene and gene-environment interactions. It is well known both that the presence of an interaction in a statistical model depends on the model being employed and furthermore that a statistical interaction need not correspond to an interaction in any biologically or physically meaningful sense. The sufficient cause framework makes reference to the actual causal mechanisms, referred to as sufficient causes, involved in bringing about the outcome. When two or more binary causes participate in the same causal mechanism, synergism is said to be present. Sometimes synergism cannot be identified from data; when data do imply that synergism must be present then a sufficient cause interaction is said to be present. The theory and methods developed through the proposed research lead to empirical tests for sufficient cause interactions and thus constitute tests for the joint presence of two or more causes in a single causal mechanism. The aims of the research are to extend the theory concerning the sufficient cause framework for dichotomous exposures, to develop theory for sufficient cause interaction for ordinal and categorical exposures, to develop multiply robust semiparametric tests for the presence of sufficient cause interactions, and to characterize those forms of exposure misclassification for which tests for sufficient cause interactions yield valid conclusions. The research will provide a set of techniques that can be used to identify mechanistic interactions in biological systems and will develop both a theoretical framework in which to conceptualize these mechanistic interactions and provide methods to empirically test for such interactions. The research will be useful in identifying mechanistic gene-gene and gene-environment interactions which could increase our understanding of genetic mechanisms. The implications of the research on sufficient cause interactions for understanding the mechanistic implications of standard gene-gene and gene-environment interaction tests and study designs will be explored and the methods developed will be applied to several data sets in the Health Effects of Arsenic Longitudinal Study. The research will make important advances to the statistical literature on the concept of interaction and on the implications of measurement error for causal inference. The overall research program will contribute to our understanding of the concepts of causation which form the foundation of the statistical literature in causal inference and which are being employed in medicine, epidemiology, psychology, genetics, computer science, philosophy, sociology, education and economics. PUBLIC HEALTH RELEVANCE: The statistical methodology developed in the proposed research will be useful in identifying mechanistic interactions in biological systems and in the analysis and interpretation of studies in genetic epidemiology of gene-gene and gene-environment interactions. The methods will be applied to data in the Health Effects of Arsenic Longitudinal Study in order to provide knowledge about the underlying pathophysiology and mechanisms by which arsenic exposure may lead to diseases. The research will make important advances to the statistical literature on the concept of interaction and on the implications of measurement error for causal inference.