Figure 6: (a) Optical flux for a s– s interval (last peak) with the approximation shown in Figure 5 subtracted. Smooth line shows the best-fit sinusoidal approximation of the data with s period. (b) Power density spectrum of this data, estimated by bootstrapping method—by generating a large number of sample time series by randomly shuffling the original light curve, what completely destroys its time structure while keeping the distribution of its values, and by studying the distribution and quantiles of resulting power densities. Horizontal lines represent mean noise level (lower) and a level of noise deviations with significance (upper), estimated by bootstrapping number of time series from the original data set. Vertical line corresponds to the period of the sinusoidal approximation shown in (a), clearly coincided with the peak of power spectrum. The probability of a random appearance of a feature like the one seen in any of 39 frequency bin is ~0.01.