Closed steven-murray closed 5 years ago
I have added the basic machinery to do this into powerbox (>=v0.5.2), and as of 03f80d67c6840e4e6a2b40d2719b5a01b68981e9, this is calculated when averaging the power. Nevertheless, I am not convinced that this will work okay. My initial tests on a non-foreground box indicate that the estimate of the variance is way off. It may be more useful to use an analytic covariance model.
Do I still need to find the covariance of the foregrounds?
We definitely need the covariance, but I'm not sure of the best way of doing it -- whether by numerically finding it from simulation, or by analytically computing it. I was hoping the former would be good enough, but it might not be.
We need to calculate the variance (or perhaps covariance) of the power along with the power itself, to be used in weighting.