A Survey Length for AGN Variability Studies

Abstract: The damped random walk (DRW) process is one of the most commonly used and simplest stochastic models to describe variability of active galactic nuclei (AGN). An AGN light curve can be converted to just two DRW model parameters - the signal decorrelation timescale $\tau$ and the asymptotic amplitude ${SF}{\infty}$. By simulation means, we have recently shown that in order to measure the decorrelation timescale accurately, the experiment or the light curve length must be at least 10 times the underlying decorrelation timescale. In this paper, we investigate the origin of this requirement and find that typical AGN light curves do not sufficiently represent the intrinsic stationary process. We simulated extremely long (10,000$\tau$) AGN light curves using DRW, and then measured the variance and the mean of short light curves spanning 1-1000$\tau$. We modeled these light curves with DRW to obtain both the signal decorrelation timescale $\tau$ and the asymptotic amplitude ${SF}{\infty}$. The variance in light curves shorter than $\approx30\tau$ is smaller than that of the input process, as estimated by both a simple calculation from the light curve and by DRW modeling. This means that while the simulated stochastic process is intrinsically stationary, short light curves do not adequately represent the stationary process. Since the variance and timescale are correlated, underestimated variances in short light curves lead to underestimated timescales as compared to the input process. It seems, that a simulated AGN light curve does not fully represent the underlying DRW process until its length reaches even $\approx30$ decorrelation timescales. Modeling short AGN light curves with DRW leads to biases in measured parameters of the model - the amplitude being too small and the timescale being too short.