On the Evaluation of Different Statistical Procedures for Microarray Data

• Main Authors: Chia-Jui Chuang, 莊佳叡
• Other Authors: Mi-Chia Ma
• Format: Others
• Language: en_US
• Published: 2004
• Online Access: http://ndltd.ncl.edu.tw/handle/09067501455846456120

碩士 === 國立成功大學 === 統計學系碩博士班 === 92 === 　　The development of microarray is very fast at the time being, but a unified data analysis mode does not exist. This research is to study grouping of genes. The first thing is to introduce the two experiments of microarray (one is fluorescence, and another is colormetry) and another thing is how to use statistics to group genes. 　　Presently, the factor analysis and cluster analysis are usually used to group genes. In this thesis, analysis of variance (ANOVA) is proposed to group genes. The advantages are: (1) It can be analyzed no matter how large the genes set is. (2) The amount of the gene will not affect the result just as factor analysis does. (3) It will not cause the factor loading to be zero because of the two same gene presentations. (4) It is not like the cluster analysis that is difficult to obtain the grouping result of the genes by dendrogam. 　　Next, the advantage and defect of different grouping methods are compared by a real data. Then, different ways are simulated to generate data, and then the incorrectness are compared among the different grouping methods by rate of erroneous judgment. Finally, we discuss the usage opportunity of different statistic methods in every different situations.