Examples of results of GP regression and amplitude spectrum fitting for synthetic light curves. The left and right columns show three examples that resemble yellow and blue subgroup members in terms of their quality factor, respectively. The blue and green shaded regions denote the 94% confidence interval for ν_char for each method. In all cases, the true input value of ν_char is recovered well by the GP model with CELERITE2, whereas the MCMC method of Bowman et al. (2020) sometimes overestimates the value of ν_char because of the more rigid profile imposed in the Fourier domain, as demonstrated by a comparison with the noiseless input SHO model.