We recently published a method to normalize data that corrects for batch effects and allows for pooling of raw data across studies. In this post, I want to discuss more about the important (but potentially confusing) aspect of our method, where we add noise to the zeros before performing the normalization. I’ve gotten a few …