Semi-parametric Methods for Multivariate Survival Data

描述(由申请者提供)：这个项目将开发和研究新的方法来分析多变量删失失效时间数据 心血管疾病、癌症和其他生物医学研究。 该提案描述了四个项目。第一个项目涉及 具有多变量失效时间的协变量效应的非参数建模 数据。将开发用于推理的半参数方法 部分线性模型和变系数模型。渐近和有限 我们将研究所建议的统计方法的样本性质。数据 防止铀矿工的癌症风险和围产期合作项目 利用提出的方法进行了分析。 第二个项目考虑了非线性的假设检验问题 广义伪偏似然比方法的协变量效应 用于测试参数与非参数以及线性与非线性 部分线性模型和变系数模型的假设 将会被开发出来。拟议程序的统计特性将是 学习。建议的方法将应用于Collaborative的数据 围产期项目，铀矿工的癌症风险，以及 协作性血脂研究诊所计划(LRC)。 第三个项目涉及多变量失效的模型选择技术 时间数据。提出了一种惩罚伪偏似然方法。 模型和变量的选择。渐近和有限样本属性将是 调查过了。协同型血脂家族研究的数据分析 研究诊所计划(LRC)和弗雷明翰心脏研究将 已执行。 第四个项目涉及多类型经常性的统计推断。 事件数据。提出了一种估计方程的方法，用于估计 多类型重现事件边际均值模型中的均值比参数 数据；关于非线性协变量的推断的半参数方法 将研究边际均值模型中的影响；并推广部分 似然比检验和惩罚伪偏似然方法将是 为这类数据进行了调查。渐近和有限样本属性将是 调查过了。左心功能不全研究的数据分析 将会被执行。

DESCRIPTION (provided by applicant): This project will develop and investigate new methodology for analyzing multivariate censored failure time data from cardiovascular disease, cancer and other biomedical research. The proposal describes four projects. The first project deals with nonparametric modeling of covariate effects with multivariate failure time data. Semi-parametric methods for inferences will be developed for the partially linear model and varying-coefficients model. Asymptotic and finite sample properties of the proposed statistical methods will be studied. Data from Cancer Risk in Uranium Miners and Collaborative Perinatal Project will be analyzed using the proposed methods. The second project considers the problem of hypothesis testing for non-linear covariate effects with Generalized pseudo-partial likelihood ratio procedure for testing parametric versus non-parametric and linear versus non-linear hypothesis for the partially linear model and the varying-coefficients model will be developed. Statistical properties of the proposed procedures will be studied. The proposed methodologies will be applied to data from Collaborative Perinatal Project, Cancer Risk in Uranium Miners, and the Family Study of the Collaborative Lipid Research Clinics Program (LRC). The third project concerns model selection techniques for multivariate failure time data. A penalized pseudo-partial likelihood method will be developed for model and variable selection. Asymptotic and finite sample properties will be investigated. Analysis of data from the Family Study of the Collaborative Lipid Research Clinics Program (LRC) and the Framingham Heart Study will be performed. The fourth project concerns statistical inferences for multi-type recurrent events data. An estimating equation approach is proposed for estimating the mean ratio parameters in marginal means models for multi-type recurrent events data; a semi-parametric method for inferences about non-linear covariates effects in the marginal means model will be studied; and generalized partial likelihood ratio tests and penalized pseudo-partial likelihood method will be investigated for such data. Asymptotic and finite sample properties will be investigated. Analysis of data from the Studies of Left Ventricular Dysfunction will be performed.