A simple correction for multiple comparisons in interval mapping genome scans
A simple correction for multiple comparisons in interval mapping genome scans
复制标题
DOI：
10.1046/j.1365-2540.2001.00901.x
复制
发表时间：
2001-07-01
期刊：
影响因子：
3.8
通讯作者：
Cheverud, JM
中科院分区：
文献类型：
--
作者：
Cheverud, JM
已经提出了几种方法来纠正间隔映射基因组扫描中使用的逐点显着性阈值。提出了一种基于Bonferroni检验的显着性阈值校正方法。该测试涉及根据观察到的标记相关矩阵的特征值的方差计算在基因组扫描中执行的独立比较的有效数量。标记的相关性越高，特征值的方差就越高，并且在染色体上进行的独立测试的数量就越少。通过在 500 个群体规模中沿不同长度和标记密度的染色体绘制 1000 个正态分布表型来评估该方法。将从模拟中获得的实验显着性阈值与使用 Bonferroni 准则和新开发的基因组扫描中独立测试有效数量测量方法计算的阈值进行比较。 Bonferroni 计算产生的显着性阈值与通过模拟获得的阈值非常相似。 Bonferroni 和模拟分析的阈值水平在很大程度上取决于标记密度和染色体大小。在 5% 和 10% 逐点显着性水平下获得的阈值存在约 1% 的轻微偏差。这里介绍的方法为多重比较提供了相对简单的校正，可以使用标准统计包轻松计算。
Several approaches have been proposed to correct point-wise significance thresholds used in interval-mapping genome scans. A method for significance threshold correction based on the Bonferroni test is presented. This test involves calculating the effective number of independent comparisons performed in a genome scan from the variance of the eigenvalues of the observed marker correlation matrix. The more highly correlated the markers, the higher the variance of the eigenvalues and the lower the number of independent tests performed on a chromosome. This approach was evaluated by mapping 1000 normally distributed phenotypes along chromosomes of varying length and marker density in a population size of 500. Experiment-wise significance thresholds obtained from the simulation are compared to those calculated using the Bonferroni criterion and the newly developed measure of the effective number of independent tests in a genome scan. The Bonferroni calculation produced significance thresholds very similar to those obtained by simulation. The threshold levels for both Bonferroni and simulation analysis depended strongly on the marker density and size of chromosomes. There was a slight bias of about 1% in the thresholds obtained at the 5% and 10% point-wise significance levels. The method introduced here provides a relatively simple correction for multiple comparisons that can be easily calculated using standard statistics packages.