Combined-survey estimation and Markov-chain simulation of childhood obesity

描述（由申请人提供）：从了解儿童肥胖和设计有效干预措施以解决这一对公共卫生的主要威胁的文献中出现了两个主要的研究需求。首先是需要共同理解生物和社会决定因素及其相互作用。第二是需要更好地了解儿童肥胖作为体重状况的纵向轨迹。为了满足这些需求，本研究将开发结合两个纵向调查数据集的方法，以估计和模拟从出生到青春期的体重状况轨迹。同时对生物和社会影响的多个（超过三个）层面进行建模，超出了当前方法和数据的可行性。单次调查可能有一些，但不是所有的变量都需要估计这些影响，单次调查也没有足够的样本量来估计纵向模型中的这些多重影响。我们对如何包含三个以上影响级别的问题的解决方案涉及：(1)使用现有的统计建模方法，在任何一个方程中最多可包含两个或三个影响级别；(2)利用一次调查到另一次调查的多重插值，允许在两次调查中汇集儿童观测结果，从而大大增加该模型估计的有效样本量；(3)使用马尔可夫链模拟模型，允许在描述儿童体重状况轨迹的四个独立方程中积累三个以上级别的影响。然后，从模拟的童年生活路径中得出生物、行为和社会经济变量以及潜在的政策干预措施对儿童肥胖的绝对和相对影响的估计。

DESCRIPTION (provided by applicant): Two major research needs emerge from the literature on understanding childhood obesity and on designing effective interventions to address this major threat to public health. The first is a need for a joint understanding of biological and social determinants and their interactions. The second is a need to develop a better understanding of childhood obesity as a longitudinal trajectory of weight status. To meet those needs this study will develop methods that combine two longitudinal survey datasets to estimate and simulate trajectories of weight status from birth to adolescence. Modeling multiple (more than three) levels of biological and social influence simultaneously is beyond the feasibility of current methods and data. A single survey may have some, but not all of the variables need to estimate these effects, nor will a single survey have sufficient sample size for estimation of these multiple effects in a longitudinal model. Our solution to the problem of how to include more than three levels of influence involves: (1) the use of existing statistical modeling methods for up to two or three levels of influence in any one equation; (2) the use of multiple imputation from one survey to another to allow pooling of child observations across two surveys to greatly increase effective sample sizes in this model estimation; and (3) the use of Markov-chain simulation models to allow for more than three levels of influences to be accumulated across four separate equations describing the childhood weight-status trajectory. Estimates of the absolute and relative impacts on childhood obesity of biological, behavioral, and socio-economic variables, and of potential policy interventions, are then derived from simulated childhood life paths. PUBLIC HEALTH RELEVANCE: In this study, we develop statistical and computational methods that combine two longitudinal survey datasets to estimate and simulate trajectories of weight status from birth to adolescence. Estimates of the impacts on childhood obesity of biological, behavioral, and socio-economic variables, and of potential policy interventions, are derived from simulated childhood life paths.