Well CMOS circuits, in fact they are nonlinear systems, but if we wanted to use as linear circuits (as an example an amplifier), we will require a technique for this propose. Exactly, the mathematical tool for this is the small signal model through the truncated first order Taylor series of the I-V characteristic of each MOS transistor. Thus for manual computation for a human being, it is easier representing a quasi-linear circuit of the real CMOS and applied our well know network theory instead of calculating the nonlinear equation of the whole transistor net, and then applying Taylor series of such equation and then getting the first term (this another way to do it without passing the small signal model but it is very difficult even for computers).