Continuous-time processes have been very much in the focus of interest recently, mainly because of high-frequency data available in various areas of applications like in finance and turbulence.
We suggest a new method for the estimation of the kernel function of a continuous-time moving average process driven by a finite variance process with uncorrelated and weakly stationary increments.
Aiming at high-frequency data we investigate the behaviour of a continuous-time model, when sampled at a fine grid, when the grid size tends to zero.
As a prominent example, we illustrate the case of CARMA processes in full detail, stressing which features can be extented to the class of processes with regularly varying spectral densities.