I assume that you are modelling some signals based on a set of parameters and variables. If so then you will most likely develop a (some) linear predictive model(s). These will be of the form G(z) = BX where:
z is a set of parameters, which are functions of the variables X; (x1, x2,....,xn) and B is a set of coefficients, which you could call linear predictor coefficients.
For the mathematical side of it might I recommend that you refer to "An Introduction to Generalized Linear Models" by Annette Dobson (Chapman and Hall, 1990)
