1-
The authors state that they used a significance threshold based on p=0.005 using Bonferonni correction for 10 independent tests per genetic model (additive x 2, dominant, recessive, per allele) for each SNP. So overall, it seems like there is 50 tests for each SNP (10 tests x 5 genetic models). Should the corrected p-value include all tests = 4 SNPs* 5 genetic models * 10 tests (e.i. p= 0.05/200 = 0.00025)? This should be verified with an expert in genetic epidemiology and biostatistics.
请教大家这个问题,我做了4个基因位点的meta分析,每个位点分了五个基因模型,而每个模型分了10个亚组,我用的巴菲尼多重比较,是按照每个基因模型来设定的显著性水平,因为每个模型有10个亚组,所以显著性水平就是0.05/10=0.005,但是reviewer 现在提了这个问题,就是把我所有位点的都综合起来,也就是4个位点*5个基因模型*10个亚组,一共是200个亚组,所以显著性水平应该是0.05/200=0.00025,但是这样做的话,我所有的分析都没有意义了。
请教大家,这个问题我该如何回答呢?非常感谢。
4- About rs17300539, it is noteworthy that the ORs seem to be higher than ORs for rs266729 in most tests reported here, but most associations were not statistically significant. The authors should discuss this observation, and offer possible explanations (only white? Smaller sample size? MAF ?) and should provide power calculations for their main analyses.
我做两个基因位点的meta分析,比如说1位点与2位点,其中2位点的有意义。现在reviewer提了一个问题,就是说他注意到1位点的or值几乎都大于2位点,但是1位点的却没有统计学意义,问我原因是什么?还让我讨论一下。