[galaxy-user] cuffdiff variance calculations

Hello forum, I was reading some of the theory behind cuffdiff and how it calculates for DE genes. My question: if I do not have replicates, how does cuffdiff do the test statistic which is dependent on the variance? The manual says "Note that in order to calculate the test statistic T, we need to know the variance of the expression level in each condition" Doesn't this mean that I am supposed to have replicates?Otherwise how can we have variance if we only have ONE SAMPLE? I would appreciate if anyone could explainthis to me. Ciffdiff caclulates DE even when I give him one sample agaist another one...how reliable is the reult? Thanks... ib