I've got an experiment with 3 samples and 2 replicates. I'm seeing weird things with the 95% confidence intervals (conf_hi and conf_lo in genes.fpkm_tracking.) For example for ENSG00000078237 we have FPKMs of 77.2135 & 0.565956 in one sample. The 95% confidence intervals are calculated as 36.8779 - 40.90160. 
There have been a few older threads about confidence intervals from cuffdiff. I think this is a new problem though - the confidence intervals are too narrow and this is Cuffdiff v2.2.1. Perhaps I'm missing something - I would be grateful if anyone can shine some light on this.
There have been a few older threads about confidence intervals from cuffdiff. I think this is a new problem though - the confidence intervals are too narrow and this is Cuffdiff v2.2.1. Perhaps I'm missing something - I would be grateful if anyone can shine some light on this.