removebatcheffect give same foldchange

Question

[skadum]

Hello everybody,

I am using the removebatcheffect function with simulated dataset using rnorm in order to test it. So i have one experimental factor with two groups and one batch effect with 2 groups. The design is complet and balanced. The correction seems to work well when i visualizing data on boxplot before and after correction but i dont understand why fold-changes between my two experimental groups before and after are the sames.

I tried the function using only the batch argument and also with the design argument but fold-changes are the sames each time.

Did anybody observe this using these function ?

Thank you.

Answer 1

[skadum]

I think i found a solution to my problem in looking removebatcheffect function in details . The explication come from the experimental design: while my batch effect is balanced through the experimental factor, so the batch estimated variance quantity that is substracted to the original data is balanced too through samples. Therefore means would not change even if values are not the same. I thought i could have bigger foldchanges in removing the batch effect but that was wrong with this experimental design.

---