Extended Data Fig. 7: Error analysis for the short-wavelength data.
From: High atmospheric metal enrichment for a Saturn-mass planet
We compute the residuals between the best-fitting model and the light curve, shift the residuals one point at a time, add the residuals back to the model (in order to preserve the time-correlated structure in the data), and find the best fit again (i.e., a “prayer bead” analysis). The histogram shows the distribution of the eclipse depths for the two visits. The results from the first visit (blue distribution) match the expectation for white noise, but the results for the second visit (green distribution) are highly non-Gaussian, indicating correlated noise in the data.
