Multivariate Models and Methods for Correlated Data in Non-Standard Settings

先进的测量工具的出现产生了新的数据收集模式，产生了非常大的复杂数据，通常涉及混合离散和连续变量或具有复杂多级相关性的高维二进制数据。 其复杂的结构通常需要非标准的统计方法，通常需要计算密集型方法。 这些在工程、金融、医药和健康领域的许多应用中特别常见。 通常依赖于数据或数据的一些适当变换遵循高斯分布的假设的常规工具不能直接应用于这些上下文中。 这里提出的研究涉及的多变量模型和方法的发展，在非标准的数据设置中的应用。 它特别关注的情况下，涉及复杂的依赖结构所产生的数据包括混合的结果，一方面，和高维相关的二进制变量，另一方面。 特别强调开发和传播新的方法，供工程和医学/健康科学从业人员使用。 拟议的研究预计将产生改进的，灵活的，强大的数据分析技术。

The advent of sophisticated tools of measurement has given rise to new modes of data collection resulting in very large complex data, oftentimes involving mixed discrete and continuous variables or high-dimensional binary data with complex multi-level correlations. Their complex structure typically requires non-standard statistical approaches that usually entail computationally intensive methodologies. These are particularly common in many applications in engineering, finance, and in medicine and health. Conventional tools that generally rely on the assumption that the data, or some suitable transformations of them, follow a Gaussian distribution, do not directly apply in these contexts. The research proposed here concerns the development of multivariate models and methodologies for application in non-standard data settings. It pays specific focus on situations involving complex dependence structures arising from data comprising mixed outcomes, on the one hand, and high-dimensional correlated binary variables, on the other. Particular emphasis is given on development and dissemination of new methodologies for use by practitioners in engineering and the medical/health sciences. The proposed research is anticipated to yield improved, flexible, and powerful techniques for data analysis.