Such functions are called linear, for reasons that I'll explain soon enough (well, we generally want another condition for linear functions involving scalar multiples of the variable, but I'll ...

We suggest reducing the design matrix to row echelon form as a means of finding the structure of all estimable functions. We illustrate the procedure with two examples.

R. Thompson, R. J. Baker, Composite Link Functions in Generalized Linear Models, Journal of the Royal Statistical Society. Series C (Applied Statistics), Vol. 30, No ...