STATISTICAL ISSUES IN AIDS RESEARCH

该项目将在以下几个主要领域采用统计方法： 艾滋病研究。 围绕艾滋病危机的紧迫感导致需要 能够更有效和高效地设计和分析 临床试验为了减少进行培训所需的时间和资源， 研究，多变量扩展组序贯设计将是 开发使用计数过程工具，并长期代理的作用 将研究长期临床疗效测量。新的鲁棒 用于误测或缺失协变量分析的非参数方法 将探索数据。 艾滋病毒流行率估计和艾滋病流行参数预测 提高对流行病动态的认识， 预测其未来公共卫生规划方向的基础 目的精确的流行率估计将通过开发 通过使用对合并样本进行的分析的统计方法 聚合酶链反应测定，并通过使用小面积估计 技术.基于伪似然的新统计方法 函数将开发的各种参数的估计 艾滋病流行，当使用艾滋病毒流行率数据时， 其他数据来源的参数估计。所有提议的方法 将在实际的艾滋病研究数据集中实施。 ARc和AIDS患者入组的临床和心理过程 卫生保健系统可能相当复杂。灵活的统计 将探讨一种方法，以便更好地了解 本课程通过分析登记册和 临床试验特别感兴趣的将是多变量点 允许时间依赖协变量的过程模型。 有特别令人信服的理由来获得方法， 确保保密的艾滋病研究和病人护理数据的安全。 将开发两种概率数据安全方法，基于 散列码识别方法与部分信息文件匹配 方法.

This project will pursue statistical methods in several main areas of AIDS research. The sense of urgency surrounding the AIDS crisis has led to the need for methods enabling more effective and efficient design and analysis of clinical trials. To reduce the time and resources required to conduct studies, multivariate extensions of group sequential designs will be developed using counting process tools, and the role surrogates for long term clinical efficacy measures will be investigated. New robust nonparametric methods for analysis of mismeasured or missing covariate data will be explored. Estimation of HIV prevalence and projection of AIDS epidemic parameters improves the understanding of the dynamics of the epidemic and provides a basis for predicting its future course for public health planning purposes. Precise estimates of prevalence will be obtained by developing statistical methods for assays performed on pooled samples, through use of Polymerase Chain Reaction Assays, and by using small area estimation techniques. Novel statistical methods based on a pseudo-likelihood function will be developed for the estimation of various parameters of the AIDS epidemic, when one uses HIV prevalence data augmented by parameter estimates from other sources of data. All proposed methods will be implemented in actual AIDS research data sets. The clinical and psychological course for ARc and AIDS patients entering the health care system can be quite complex. Flexible statistical methodology will be explored which will allow a better understanding of this course through the analysis of data arising in registries and in clinical trails. Of particular interest will be multivariate point process models which allow time dependent covariates. There are particularly compelling reasons to obtain approaches to insuring security of confidential AIDS research and patient care data. Two probabilistic data security approaches will be developed, based on hash code identifier methods and partial information file matching methods.