Top: correct 16 kHz posterior at the reference time in [2] computed by ringdown (solid) and pyring (dashed), contrasted with the incorrect posterior in [1] (dotted), as in the second panel of Fig. 4 in the Supplemental Material of [1]. Bottom: besides failing to target , results in [1] suffered from a bug that mislabeled LIGO time stamps and analyzed insufficient data (length ); the bottom panel shows the effect of fixing each of these issues, to eventually yield the true posterior at (thick green), which differs only slightly from the intended target at (top blue).
posterior median (line) and 68% HPD (band) for different fit times relative to (abscissa). Top: GW150914 analyzed at 2 kHz (blue) and 16 kHz (gray), compared to the result in [4] (dots); a dashed line marks the time used in [1] and incorrectly labeled as (solid line); a dot-dashed line shows the peak time estimate in [1]. Bottom: simulation of two-damped sinusoids drawn from a GW150914 posterior without noise (green) and in synthetic LIGO noise (gray) at 16 kHz; to avoid an abrupt start, the injection is a “ring-up” before , as in Fig. 10 in [8]. The GW150914 analyses prefer the overtone over a broad set of times around the inferred peak (top), and the posterior tracks the decay of the mode plus, for the 16 kHz runs, additional jitter from high frequency noise, consistent with damped sinusoids in Gaussian noise (bottom).