If so, then you measure Ct values (or dCt, if you normalize to some reference). The difference of such (d)Ct values is a log2 ratio of the initial (normalized) amounts of the target sequence.
Thus, all you need is to fit a usual normal linear model that will give you the differences (model coefficients) together with their standard errors. Very simple.
--
*(that would actually be "qrtPCR", not to confuse rt ["real-time"] with RT ["reverse-transcription"]).