Psychometric and Genetic Assessments of Substance Use

该项目旨在开发一系列新的方法来确定药物使用和滥用的表型。这位将军 方案是从理论上开发统计模型，在用户友好的软件中实现它们，并进行检验 它们的统计特性。那些表现足够好的模型将应用于一组或多组 为评估物质使用情况带来新的见解的数据。第一个目标是将模型扩展到 阶乘不变性，这是检验组间差异的基础。主分机 将不仅允许测试不同组之间的不变性，而且还允许测试不同组内的不变性 考虑到连续的变量，如年龄。这一方法将被应用于验证性因素分析 潜在类别分析，以及代表因素和潜在类别的混合的模型，并将被 能够处理二进制、有序和连续的观察变量。这种方法应该被证明在 评估人群中的药物滥用模式是否代表责任的持续变化或 是否存在不同的群体。第二个目标是在以下背景下扩展政权转换的模型 增长曲线等因素混合模型。此目标旨在为以下数据提供更好的模型 涉及物质使用的开始和抵消，并帮助发现异质性。三是大力发展 用于分析某些形式的部分匿名数据的方法，例如涉及随机化的数据 回应。将比较这些方法在检测与预测器的关系方面的性能， 部分随机化数据的后遗症和相关性，包括亲属和 结果。所有模型开发的设计都将允许分析和利用收集到的数据 来自亲属的数据，将包括遗传标记数据的模型，用于连锁和关联研究。 应用的数据分析将产生实质性的结果，指导模型开发，并测试稳健性。一个 将汇集和分析横截面、纵向和遗传信息数据集的阵列。

This project aims to develop a series of novel approaches to phenotyping drug use and abuse. The general scheme is to develop statistical models from theory, implement them in user friendly software, and examine their statistical properties. Those models that perform sufficiently well will be applied to one or more sets of data to bring new insight into the assessment of substance use. The first goal is to extend of models for factorial invariance, which form the basis of testing for differences between groups. The primary extension will be to allow testing of invariance not merely between distinct groups, but also within groups that vary with respect to continuous variables such as age. This approach will be applied to confirmatory factor analysis, tc latent class analysis, and to models that represent mixtures of both factors and latent classes, and will be able to handle binary, ordinal and continuous observed variables. The method should prove valuable in assessing whether substance abuse patterns in the population represent continuous variation in liability or whether distinct groups exist. The second goal is to extend models for regime switching in the context of growth curve and other factor mixture models. This aim is intended to provide a better model for data that involve onset and offset of substance use, and to assist in uncovering heterogeneity. Third, we will develop methods for the analysis of certain forms of partially anonymized data such as those involving randomized response. These methods will be compared for their performance at detecting relationships with predictors, sequelae and correlates of partially randomized data, including resemblance between relatives and outcomes. All model development will be designed to permit the analysis and exploitation of data collected from relatives, and will include models for data on genetic markers, for both linkage and association studies. Applied data analyses will yield substantive results, guide model development, and test for robustness. An array of cross-sectional, longitudinal, and genetically informative datasets will be assembled and analyzed.