Hi, I am trying to learn how to use SVA. The example given in the bioconductor manual runs pretty smoothly with 1000 genes and 20 samples (10 in each group, genes 1-300 respond to the primary variable of interest, genes 200-500 respond to some other variable). However, if I change the matrix to either the following two scenarios: A. There is no difference between the two groups. svadata <- cbind(matrix(rnorm(10000),nc = 10),matrix(rnorm(10000),nc = 10)) B. There is a big difference between the two groups. svadata <- cbind(matrix(rnorm(10000),nc = 10),matrix(rnorm(10000)+2,nc = 10)) The run returns "No Significant surrogate variables". I am wondering under what conditions can SVA be applied? Must it be a mixture of both differentially expressed genes and non differentially expressed genes? The data at my work has many confounding variables and the p-value distribution is tailed towards 0, does SVA apply to this scenario? Thanks a bunch, -Jack [[alternative HTML version deleted]]