I carried out a screen for 400 different genes. This was done in batches of 20 genes. Each batch includes 2 replicas of an internal control: a gene that works as a negative control. There is batch to batch variation and the differences between the batch means for the parameter being measured are statistically significant. In addition, the differences between the internal controls in the different batches are also statistically significant between some batches.

How do I measure the extent of the batch to batch variation that is due to noise from the extent that is due to "true" variation"?

Thank you. J