Semi-Parametric Survival Methods in Dental Research

描述（由申请人提供）：在面向患者的牙科研究中，由时间到事件、受试者内聚类观察组成的数据集很常见，例如，从每个患者的多颗牙齿甚至每颗牙齿的多个表面记录龋齿数据。对每位患者的多个观察数据进行分析是一个具有挑战性的，但往往被忽视的命题。大多数常规使用的统计分析程序都需要一个共同的假设，即所有的观测结果都是独立的。在存在聚类观察的情况下，这些程序通常不能直接适用。为了做出有效的推断，需要更复杂的统计工具来考虑集群内部的相关性。复杂的生存方法分析在面向患者的牙科研究中的应用是有限的。为了解决这一不足，庄博士（首席研究员）正在提交R03奖，以资助他在以患者为导向的牙科研究的定量分析方面的特殊兴趣和技能。该奖项的广泛目标是促进首席研究员在牙科研究中应用健全和严格的生存分析方法，并在未来成为独立的以患者为导向的牙科研究人员、导师和临床医生。庄博士的环境非常适合实施以患者为导向的研究。该部门和机构致力于为个人提供以患者为导向的研究方法。R03奖将为受保护时间内至少35%的研究工作提供资金。R03奖项的第一个具体目标是通过边缘方法通过广义线性转换模型确定与种植体失败和并发症相关的暴露。第二个具体目标是通过脆性方法确定与种植体失败和并发症相关的暴露，并通过广义线性转换模型对两者进行比较和对比

DESCRIPTION (provided by applicant): Datasets composed of time-to-event, within-subject clustered observations are common in patient oriented dental research, and e.g. caries data are recorded from multiple teeth or even multiple surfaces per tooth from each patient. The analysis of data composed of multiple observations per patient is a challenging, yet often overlooked, proposition. Most routinely used statistical analysis procedures require the common assumption that all observations are independent. In the presence of clustered observations, these procedures are usually not directly applicable. To make valid inferences, more sophisticated statistical tools are needed in order to take within cluster correlations into account. The application of sophisticated survival method analyses in patient-oriented dental research is limited. To address this deficiency, Dr. Chuang (principal investigator) is submitting a R03 award to fund his special interests and skills in the quantitative analyses of patient-oriented dental research. This award's broad goal is to facilitate the principal investigator to develop sound and rigorous survival analytic methods with applications in dental research and in the future to function as an independent patient-oriented dental researcher, mentor and clinician. Dr. Chuang's environment is outstanding for implementing patient-oriented research. The department and institution is committed to provide individuals in patient-oriented research methods. The R03 award will provide funding for protected time with at least 35% research effort. The first specific aim of this R03 award is to identify exposures associated with implant failures and complications with a generalized linear transformation model via the marginal approach. The second specific aim is to identify exposures associated with implant failures and complications with a generalized linear transformation model via the frailty approach and to compare and contrast these two different approaches. The principal investigator has chosen this research experience because of its clinical relevance and challenging methodological and analytic issues. The analyses focus on the survival methodological issues of correlated or clustered observations with applications to dental research. The specific analytic issues addressed are generalizable to patient-oriented dental research as many studies using multiple, correlated observations per patient, i.e. probing depths, tooth loss.