Good morning,
I am currently trying to analyse a small-RNA sequencing dataset using limma package and the voomLmFit function. However, I am experiencing some issues with the p-value distribution, with some of the comparisons I am testing showing odd p-value distribution. This probably indicates that there is some kind of batch effect that I am not correcting for, even though there is no separation of the samples in the PCA plot based on any of the potential confounding variables.
I have tried to include different variables in the model to see if the p-value distribution improves, but it doesn´t. However, including some variables (for examples RNA_extraction_batch) improves the Student’s t Q-Q plot (that I read it´s good to check?). Do you have any suggestion on why this is the case and if I should use a model that improves the Q-Q plot even though it doesn´t really improve the p-value distribution? I have included the code and two examples of qq plots of the fit2$t with or without including the variables in the model.
Thank you for the help!! qq plot of fit2$t when I include additional variables in the model qq plot of fit2$t when I don´t include extra variables in the model
