Statistical Models with High-Dimensional Predictors

项目6涉及开发将适用于许多非常高的 作为CONTE中心的一部分收集的维度数据集。尤其是，我们将重点关注 具有单一结果变量和非常高维预测因子的模型，例如使用基因表达 区分自杀未遂者和抑郁的非未遂者的数据，或使用脑成像数据 来预测病人对抑郁症治疗的反应。该方法将采用强大的新技术 开发统计概念和工具，包括函数数据分析方法、机器学习 技术和预筛选算法。重点将放在开发既能实现 准确的预测并提供稳定的可解释模型，使您能够更深入地了解 自杀行为和精神疾病的生物学基础。该项目涉及开发适当的 方法，既适用于现有数据集，也适用于将作为COTE一部分收集的数据集 并使用仿真研究和实际数据对各种建模策略进行了比较 验证。

Project 6 involves developing statistical methodology that will be applicable to many of the very high dimensional datasets that are being gathered as part of the Conte Center. In particular, we will focus on models with single outcome variables and very high-dimensional predictors, e.g., using gene expression data to discriminate between suicide attempters and depressed non attempters, or using brain imaging data to predict a patient's response to treatment for depression. This methodology will employ powerful newly developing statistical concepts and tools including functional data analytic methods, machine learning techniques, and prescreening algorithms. Emphasis will be on developing models that can both achieve accurate predictions and provide stable interpretable models, allowing for a deeper understanding of the biological basis of suicidal behavior and mental illness. The project involves development of appropriate methodology, application both to existing datasets and to those that will be gathered as part of the Conte Center, and comparison among the various modeling strategies using both simulation studies and real data validations.