pairwise comparison regardless effect

Hi there,

I want to do a pairwise comparison of gene expressions (continuous values) before and after treatment of a group of patients regardless the direction of the effect. So the genes can either show either up- or down-regulation in course of the treatment. Using normal t-statistics does not help here because up-regulation "neutralises" down-regulation and vice versa. That is why no pairwise group comparison can be made.

So how to test the null-hypothesis of "there is no change of gene expression"?

Please see the attached screenshot visualising the data situation.

Many thanks!


TS Contributor
Are you looking for something like a test for "H0: The mean absolute change in gene expression is exactely 0 in the population"? You need not test that, you just need 1 case with absolute change > 0 to reject the null. And you have more than 1 case with abslute change > 0 here.

With kind regards


Your patient are paired, so could you substract the "after" result to the "before" result for each patient, plot the resulting values and analyse the distribution ?