MEASUREMENT ERROR MODELS, NUTRITION AND BREAST CANCER

描述（来自申请人的摘要）：这个持续的项目涉及 一般回归中测量误差问题的进一步研究 问题 开发的方法将应用于一些数据 调查人员所拥有的设备 这项研究将 执行的福尔斯分为三大类。 1)饮食摄入数据分析。 将FFQ与 将对常规摄入量进行调查。 缺失数据和最优半参数 方法将被讨论。 估计通常摄入量的分布将 被研究。 一种新的方法，称为非参数校准，将 为将饮食工具与未知的通常摄入量相关联而开发的 协变量的函数，如体重指数。 2)功能和结构非线性测量误差建模。 新 将发展一种非线性测量误差的一般方法 当没有对潜在分布做出假设时， 容易出错的预测器 3)半参数和降维方法： 问题驱动 在测量误差和缺失数据方面，研究人员将制定一个 半参数插入估计的一般理论，其中 半参数函数的估计和替代成一个无偏的 估计方程 在广义部分线性单指标下 模型（GPLSIM）系列，结果将从准似然模型扩展 一般的回归问题。 这个扩展的一个重要应用 将是膳食摄入研究中的非参数校准。

DESCRIPTION (From applicant s abstract): This continuing project concerns further work on the problem of measurement error in general regression problems. The developed methodology will be applied to a number of data sets that the investigators have in their possession. The research to be performed falls into three broad categories. 1) Analysis of dietary intake data. Calibration studies that relate FFQs to usual intake will be investigated. Missing data and optimal semiparametric methods will be discussed. Estimating the distribution of usual intake will be studied. A new method, called nonparametric calibration, will be developed for relating dietary instruments to usual intake as an unknown function of covariates such as body mass index. 2) Functional and structural nonlinear measurement error modeling. New general methodology will be developed for nonlinear measurement error modeling when no assumptions are made about the underlying distribution of the error-prone predictor. 3) Semiparametric and dimension reduction methods.: Motivated by problems in measurement error and missing data, the investigators will develop a general theory of semiparametric plug-in estimation, in which a semiparametric function is estimated and substituted into an unbiased estimating equation. Within the Generalized Partially Linear Single-Index Model (GPLSIM) family, results will be extended from quasi-likelihood models to general regression problems. An important application of this extension will be to nonparametric calibration in dietary intake studies.