Here, again, we follow the method of PRH92 to determine the covariance model for our continuum data. At this stage, we make the assumption that the underlying process is statistically stationary so that
where 
is the autocorrelation function that we have to determine.
This function is related to the first order structure function 
by
and can be approximated by forming pair-wise estimates for all
distinct pairs of data points in the continuum light curve, and then binning
by the time lag of the pairs. We find that the analytic form
fits the structure functions of this paper well. In fact, the reconstruction is fairly insensitive to the exact analytic form used to approximate the structure function.
Making the reasonable assumption that 
as 
, our final expression for the
autocorrelation function is
