SCORE: Principal Differential Analysis with Covariates for Functional Data

描述(申请人提供)：本建议重点扩展主差分分析(PDA)用于曲线数据分析。听觉研究中的几个应用促进了这种方法的扩展，允许协变量调整。目前，所有已发表的听觉脑干反应曲线(ABR)和皮层听觉诱发电位曲线(CAEP)的分析方法都描述了曲线的特征，如显著波形的位置和幅度。建议的统计方法将允许使用整个ABR和CAEP曲线而不是仅选择的特征进行分析，以帮助理解与年龄相关的、多发性硬化症或糖尿病相关的听觉刺激时间处理的变化，并更好地预测人工耳蜗植入的临床结果。这些方法有可能用于创伤性脑损伤的早期检测。在伊拉克和阿富汗持续部署军事人员的这段时间里，了解噪声导致的听力损失和脑外伤之间的相互作用至关重要，这是暴露在爆炸中的伊拉克自由行动/持久自由行动士兵常见的障碍(迈耶斯、威尔明顿、加伦和亨利2009年；约旦、李和赫尔弗2009年)。PDA是一种通过首先估计接近于消除噪声的曲线数据的线性微分算子来获得曲线的低维表示的方法。在具体目标(1)中，主差分分析将扩展到包括使用局部线性平滑的协变量平差。我们将解析地研究系数函数的非参数估计的渐近偏差和方差性质。偏差和方差的渐近表达式将被用来提出基于数据的平滑参数选择方法。提出了协变量重要性的检验统计量。在具体目标(2)中，开发的方法将在sPlus或R中实施，并公开提供。计算机模拟研究将验证系数估计量的偏差和方差的渐近理论对于指导平滑参数的选择和研究所提出的检验统计量的性质是有用的。在分析理解扩展的PDA的特性并实施扩展的PDA软件之后，将在咨询听觉研究人员的情况下，针对特定的目标(3)和(4)分析听觉研究数据。PI有发展的、学术的和专业的目标。学术目标是发展听觉研究的能力，以便推进解决该领域数据分析问题的新的统计方法。专业目标是在UTEP建立一个强大的统计学研究环境：让新的计算科学博士项目的博士生和统计学的硕士学生参与进来；在研究会议上发表演讲；以及与听觉研究方面的国家专家建立合作关系。发展目标的进展将通过在她的指导下统计和最近批准的UTEP计算科学计划内的学生论文发表率和数量的增加来衡量。1 这些方法有可能用于创伤性脑损伤的早期检测。

DESCRIPTION (provided by applicant): This proposal focuses on extending principal differential analysis (PDA) for analysis of curve data. Several applications in auditory research are motivating this methodological extension, which allows for covariate adjustment. Currently, all published methods for analysis of auditory brainstem response curves (ABR) and cortical auditory evoked potential curves (CAEP) describe features of the curve, such as location and amplitude of prominent waveforms. The proposed statistical methods will allow for an analysis that uses the entire ABR and CAEP curves, rather than just selected features, to contribute to the understanding of age-related, multiple sclerosis- or diabetes-related changes in temporal processing of auditory stimuli, and better prediction of clinical outcomes from cochlear implantation. The methods have the potential to be used for the early detection of traumatic brain injury (TBI). Understanding the interaction of noise induced hearing loss and TBI, disorders common to Operation Iraqui Freedom/Operation Enduring Freedom soldiers exposed to explosive blasts (Meyers, Wilmington, Gallun, and Henry 2009; Jordan, Lee, and Helfer 2009), is vital during this time of ongoing deployment of military personnel in Iraq and Afghanistan. PDA is a method for obtaining a low-dimensional representation of a curve by first estimating a linear differential operator that comes close to annihilating the noisy curve data. In Specific Aim (1), principal differential analysis will be extended to include covariate adjustments using local linear smoothing. The asymptotic bias and variance properties of the nonparametric estimators of the coefficient functions will be investigated analytically. The asymptotic expressions for bias and variance will be used to propose data-based methods for smoothing parameter selection. Test statistics for the significance of covariates are proposed. In Specific Aim (2), the methods developed will be implemented in one of Splus or R and made publicly available. Computer simulation studies will be executed to verify that the asymptotic theory for the bias and variance of the estimators of the coefficients is useful in guiding the smoothing parameter selection, and for studying the properties of the proposed test statistics. After the properties of the extended PDA are understood analytically and the extended PDA software is implemented, auditory research data will be analyzed for Specific Aims (3) and (4) in consultation with auditory researchers. The PI has developmental scholarly and professional objectives. The scholarly objective is to develop competency in auditory research in order to advance novel statistical methodologies that solve data analysis problems in the field. The professional objective is to build a strong research environment in statistics at UTEP by: involving doctoral students from the new Computational Sciences Ph.D. program and masters students in statistics; presenting at research conferences; and establishing collaborations with national experts in auditory research. Progress on developmental objectives will be measured in terms of an increased publication rate and numbers of student theses under her direction in statistics and within the recently approved Computational Sciences Program at UTEP. 1 PUBLIC HEALTH RELEVANCE: Narrative The proposed statistical methods will allow for a statistical analysis that uses the entire auditory brainstem response (ABR) and cortical auditory evoked potential (CAEP) curves, rather than just selected features, to contribute to the understanding of age-related changes, multiple sclerosis- or diabetes-related changes in temporal processing of auditory stimuli, and better prediction of clinical outcomes from cochlear implantation. The methods have the potential to be used for the early detection of traumatic brain injury (TBI). Understanding the interaction of noise induced hearing loss and TBI, disorders common to Operation Iraqui Freedom/Operation Enduring Freedom soldiers exposed to explosive blasts (Meyers, Wilmington, Gallun, and Henry 2009; Jordan, Lee, and Helfer 2009), is vital during this time of ongoing deployment of military personnel in Iraq and Afghanistan.