A Note on Modelling Seasonal Processes in Continuous Time (Q4956022)

This paper focused on stochastic differential equations of the type \[ a(D)x(t)=\varepsilon(t),\;a(d)=D^q+a_{q-1}D^{q-1}+a_1D+a_0, \] where \(D\) denotes the mean square differential operator. The coefficients \(a_0,\dots,a_{q-1}\) are real valued and \(\varepsilon(t)\) has constant spectrum \(f(\lambda)=\sigma^2/2\pi\). It is shown that such equations are capable to generate a stationary seasonal processes but not processes that are integrated of order 1 at frequency \(\pi\) (whose spectral density has a pole of order 2 at \(\pi\)). Alternatively, mixed differential-difference equations may be used for modelling seasonally integrated processes.