Validation of the 3SD approach for a noise n composed of granulation plus WGN and d = 0 in Eq. (1). We investigate three temporal sampling grids, from left to right: regular, regular with a central gap, and random. Top panels: relation p_v(t) for the detection procedure using a genuine NTS based on GOLF solar data (green; oracle). The B = 100 estimates produced by the 3SD procedure using the MHD-based NTS (gray). Sample mean value of the gray curves (black). The spectral window of each sampling grid is shown in the inset panels. In panel (a), the red line represents the analytical expression of p_v (see Eq. (13) of Sulis et al. 2020). The dotted lines indicate the test statistics for which equals a mean p-value of 1%. Middle panels: empirical distribution of at t and 90% intervals (dotted lines) around the mean (solid black lines). P-values obtained with the oracle are shown with green vertical lines. Bottom panels: for one of the B loops of algorithm 2, we show an example of the empirical distribution of the b = 10⁴ random values of the frequency index where the largest periodogram peak was found (gray). Empirical distribution of the Max test related to the oracle procedure (green; only contours are shown for clarity). The same distribution, but for a nonstandardized LSP is shown in blue (i.e., the Max test is applied to p instead of ).