Fitting a GLM is an iterative process, and so you need to define a stopping criteria. We look for relative changes to the likelihood to define convergence. Sometimes the GLM doesn't converge for some rows when the counts are mostly zero and with very sparse model matrices (although in the recent releases, typically all rows converge). You can likely ignore this message, or remove these rows from the results if you like: res = res[mcols(dds)$fullBetaConv,]
What package version are you using: packageVersion("DESeq2")?
What package version are you using: packageVersion("DESeq2")?
Comment