A common phenomenon of active galactic nuclei, which presumably harbor
supermassive black holes with masses of 
(Rees 1984), is the strong variability which can be observed
in X-ray lightcurves. These AGN lightcurves seem to show
featureless "red noise", i.e. scale-free, divergent variability at
low frequencies, often also described as flickering or 1/f fluctuation
(Lawrence et al. 1987). The 1/f term describes the power law
distribution of the spectral power with the function 
in
the power spectrum, often denoted as "1/f" behavior.
We present an alternative model to analyse the variability seen in the X-ray lightcurves of AGN. The standard method of analyzing time series in the frequency domain is discussed briefly in Sect. 2. The alternative is known as a Linear State Space Model (LSSM) based on the theory of autoregressive processes (Scargle 1981; Honerkamp 1993) which usually cannot be observed directly since the observational noise (i.e. detectors, particle background) overlays the process powering the AGN. A LSSM fit applied to the time series data yields the dynamical parameters of the underlying stochastic process. These parameters should be strongly correlated to the physical properties of the emission process. The corresponding LSSM power spectrum exhibits both the decrease of power at medium frequencies and a limitation of spectral power at low frequencies. The detailed mathematical background of LSSM and the fit procedure are described in Sects. 3 and 4. Finally we present first results using this technique with EXOSAT data from the Seyfert galaxy NGC 5506 in Sect. 5.