@dereneaton @isaacovercast Found a possible bug, using 5.10. After running through s6, I then branched a subset of samples that are all from one sequencing run and reran s456. My thought process was I wanted to recalculate the error rate/heterozygosity for this subset as it was sequenced on a different Illumina version. This ended up rewriting the original s4_joint_estimate.txt with the reduced number of samples, instead of creating a new s4_joint_estimate.txt. Should I not redo s4 in this case, and ipyrad will just recalculate average error rate/heterozygosity for a branch in step 5 that has a different number of samples?