如果set p < 0.05 as significant difference. 那么p < 0.05 和 < 0.01有区别么? 好像听说这两个没有区别。
u*8
3 楼
最好不要去枫叶国续签h1b,貌似被check很麻烦的
o*y
4 楼
the lowest probability gives much stronger evidence for rejecting the hypothesis. Sometimes, depending on the hypothesis being tested, a researcher may decide that the “less than 5%” significance level is too risky.
【在 s**u 的大作中提到】 : 如果set p < 0.05 as significant difference. 那么p < 0.05 和 < 0.01有区别么? : 好像听说这两个没有区别。
difference 的大小是看两个样品的数学期望的差值。 而这个差值在统计意义上的可靠性,是由 p-value 来评估的,叫做statistical signi ficance. 同一组数据,衡量不同的统计方法的 power 也是看 p-value,p-value 越小,这个统计 方法就越 powerful。 而天下没有免费的午餐。统计方法的 power 来自于更多更苛刻的 assumption,也即来 自于 a prori knowledge,所以往往越 powerful 的统计方法的适用面越窄。 而且每种统计方法,在使用时,都要检验其 assumptions 是否满足。例如,样本数目比 较小时,无法检验样本是否符合 t 分布,所以用 t-test 就不靠谱,而应使用 non-pa rametric tests。
difference
【在 s**u 的大作中提到】 : does it mean that the one with lower P values is more significant difference : (larger difference)than the one with higher P values?
m*7
16 楼
P means probability, it has nothing to do with how big the difference ( between control and treated samples) is. Usually in my figures, I put one star (*) on top of a bracket to indicate p< 0.05, two stars (**) to indicate p<0.01.
Most often, you would like to use both the statistical significance and the effect size. For example, in microarray analysis, it would be better to find genes with q<0.05 and fold>2. The larger your sample size is, the most significance you will get, but the effect size will stay the same.