本文選題：薈萃分析 + M-H算法　； 參考：《哈爾濱工業(yè)大學(xué)》2017年碩士論文

【摘要】：薈萃分析是將相同的多個研究的統(tǒng)計量進(jìn)行合并,概括主要研究結(jié)果,得出簡單而重要的結(jié)論。本質(zhì)上是利用統(tǒng)計方法來增強主要研究效應(yīng)的方法。本文介紹了薈萃分析方法的概念以及兩種薈萃分析模型,即固定效應(yīng)模型與隨機效應(yīng)模型,著重考慮了相應(yīng)模型的算法:M-H算法和DSL算法,并對兩種算法進(jìn)行了分析。在原算法的基礎(chǔ)上考慮權(quán)重因子的大小以及對數(shù)對原算法的影響,改進(jìn)了兩種算法。具體的,在DSL校正因子的啟發(fā)下,將M-H和DSL算法中權(quán)重設(shè)置為調(diào)和平均數(shù),運用統(tǒng)計學(xué)上的Z檢驗方法進(jìn)行檢驗,對改進(jìn)前后的算法結(jié)果進(jìn)行了比較。在應(yīng)用方面,用改進(jìn)后的薈萃分析模型算法分析高血壓遺傳變異和SNP的關(guān)系。在Pub Med和Web of Science文獻(xiàn)搜索引擎上搜索與高血壓和rs1799983相關(guān)的文章,收集整理數(shù)據(jù),進(jìn)行數(shù)據(jù)預(yù)處理,用改進(jìn)的薈萃分析模型算法對高血壓遺傳變異與e NOS基因上的rs1799983位點的關(guān)系進(jìn)行實驗與分析。通過發(fā)表偏倚分析、異質(zhì)性分析得出隱性模型雖然存在了發(fā)表性偏倚但不存在異質(zhì)性,其他模型不存在發(fā)表性偏倚但存在異質(zhì)性,同時通過Z檢驗對改進(jìn)前后的算法結(jié)果進(jìn)行比較,得出改進(jìn)后結(jié)果更明顯。
[Abstract]:Meta-analysis is to combine the statistics of the same multiple studies, summarize the main results and draw simple and important conclusions. In essence, statistical methods are used to enhance the main research effects. This paper introduces the concept of meta-analysis method and two meta-analysis models, that is, fixed effect model and stochastic effect model. The algorithms of the corresponding model, namely, the algorithm of: M-H and the algorithm of DSL, are emphatically considered, and the two algorithms are analyzed. On the basis of the original algorithm, the influence of the weight factor and the logarithm on the original algorithm is considered, and two algorithms are improved. Specifically, under the inspiration of DSL correction factor, the weights of M-H and DSL algorithms are set to harmonic average, and the results of the improved algorithm are compared by using the Z test method in statistics. In application, the improved meta-analysis model algorithm is used to analyze the relationship between hypertension genetic variation and SNP. Search for articles related to hypertension and rs1799983 on the search engine of Pub Med and web of science literature, collect and organize data, carry out data preprocessing, The relationship between the genetic variation of hypertension and the rs1799983 locus on the Enos gene was studied by using the improved meta-analysis model algorithm. Through the analysis of publication bias, heterogeneity analysis shows that although there is publication bias, there is no heterogeneity in the recessive model, and there is no publication bias but heterogeneity in other models. At the same time, the results of the improved algorithm are compared by Z test, and the improved results are more obvious.
【學(xué)位授予單位】：哈爾濱工業(yè)大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2017
【分類號】：R544.1
，

本文編號：2071381