LIGO Document T060089-v1
- Is it a systematic or a statistical error?
The review of our S3 stochastic search paper identified the fact that our point estimate varies by as much as ~0.25 sigma depending on how wide the notched mask is that removes harmonics of 16 and 60 Hz. At first blush, 20%-25% differences in answers when so few bins are affected may seem surprisingly large; however it can be shown that this is consistent with the statistics of the Gaussian noise. I attach a note analyzing a toy problem that quantifies the issue. I conclude that the fluctuations are within expectations.
DCC Version 3.4.2, contact