荟萃分析中用概括统计量与个体数据的等效性

发布时间：2018-05-30 18:46

本文选题：信息矩阵 + 线性回归　；参考：《大连理工大学》2015年硕士论文

【摘要】：荟萃分析,也叫元分析,是汇总多项具有同一研究目的、并且相互独立的研究结果,它是利用统计学科的一些方法来进行系统性的、综合性的、定量性的分析以及总结评价的一种研究方法。通过该方法可以达到提高统计的效能并且改善原来研究质量的目的。荟萃分析广泛地使用在综合多项研究方面,它是建立在流行病学和循证医学基础上的一种新的研究方法。尽管就传统上而言,荟萃分析通过结合相关研究的概括性数据进行,但是先进的技术已经使得接近各个参与者的原始数据越来越具有可行性。在本文中,我们调查了分析原始数据对结合概括统计量的相对效率。展示了对于通常使用的参数和半参数模型而言,如果主要利益参数在研究中有共同值、讨厌参数在研究中有明显的不同值、概括统计量是通过极大似然来估计的话,那么通过分析原始数据,便不会有效益增益。当主要的利益参数在各个研究中有不同值,或者各研究中讨厌参数有共同值时,我们也评估了两种方法的相对效益。我们进行了模拟研究,证实了理论结果并提供遗传学关联研究中的实证数据。
[Abstract]:Meta-analysis, also known as meta-analysis, is a collection of research results that have the same purpose and are independent of each other. They are systematic and comprehensive using statistical methods. Quantitative analysis and a research method of summing up evaluation. This method can improve the efficiency of statistics and improve the quality of the original research. Meta-analysis is widely used in a wide range of studies and is a new approach based on epidemiology and evidence-based medicine. Although traditionally meta-analysis is done by combining general data from related studies, advanced technology has made it more feasible to access raw data for each participant. In this paper, we investigate the relative efficiency of analyzing raw data in combination with generalized statistics. It is shown that for commonly used parametric and semi-parametric models, if the main benefit parameter has a common value in the study and the dislike parameter has distinct values in the study, the generalized statistics are estimated by maximum likelihood. Then by analyzing the raw data, there is no benefit gain. We also evaluate the relative benefits of the two methods when the main benefit parameters have different values in each study or when the disgust parameters in each study have a common value. We conducted simulation studies, confirmed the theoretical results and provided empirical data from genetic association studies.
【学位授予单位】：大连理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：O212.1;C81

【共引文献】