Power Spectral Density in Parameter Estimation
There are 3 PSD options in Parameter Estimation.
- PSD uploaded by detection pipeline (Pipeline PSD)
This is the best option for low-latency use but currently it should not be used since it seems to bias PE results for some reasons. - PSD estimated by LALInference (LI PSD)
It is fast but need long enough data. However, it might be possible to shorten the required data by estimating PSD with low frequency resolution and interpolating it. - Bayeswave PSD
It can only need on-source data but can be slow.
I think we would like to use Pipeline PSD, and move to LI PSD if it does not exist, and move to Bayewave PSD if available data is not long enough. Then we need to do the following tasks.
-
Make it possible to use Pipeline PSD in LALInference Parameter Estimation -
See if it is possible to shorten required data to estimate LI PSD -
Determine the best settings for Bayeswave PSD -
Make it possible for GWCelery to use Bayeswave PSD
cc: @vivien