A number of techniques have recently appeared in the literature utiliz
ing instantaneous nonlinear transformations of stochastic processes. A
s in economics most variables appear to be nonstationary, it is intere
sting to investigate the temporal properties of nonlinear transformati
ons of these processes. Using a unified approach based on Hermite poly
nomials, this paper provides some insight for polynomial, exponential,
periodic, and neural-network transformations of first-order integrate
d processes.