search

CHIMEFRB / fitburst

An open-source package of utilities for direct modeling of radio dynamic spectra.
https://chimefrb.github.io/fitburst/
MIT License
11 stars 2 forks source link

Add estimate on amplitude #54

Closed ketansand closed 1 year ago

ketansand commented 2 years ago

Some burst are not fitting correct amplitude. We can provide the guess from our mcmc fits. This affecting very bright events.

Screen Shot 2022-03-21 at 3 27 57 PM
DanielaBreitman commented 2 years ago

I tried adding the amplitude in the initial guess and the fit was not improved.

ketansand commented 2 years ago

We can try with Fitburst_example_generic.py with preprocessing enabled as shared by Emmanuel in this slack chat - https://chime-experiment.slack.com/archives/C038CUXAH27/p1648067677124639

ketansand commented 2 years ago

Use preprocessing of data reader to adjust baseline.

DanielaBreitman commented 2 years ago

@emmanuelfonseca can you confirm that using the preprocessing step helped get rid of this issue? If I look in the fitburst_example_generic.py on the main branch, it seems like it is not used actually (https://github.com/CHIMEFRB/fitburst/blob/2ed613aa862f12d700c19452d1fb7ab8d7543579/fitburst/pipelines/fitburst_example_generic.py#L331)

DanielaBreitman commented 2 years ago

This issue is indeed fixed with @emmanuelfonseca 's suggestion to change the variance_range to [0., 1.]. The fix is then that I can add a flag: preprocess = True / False that can be passed by command line that will do the pre-processing step OR we can just do the pre-processing every time. Note that one issue here is that the floor is off for some reason. image

emmanuelfonseca commented 1 year ago

i believe this has been addressed: this is partly due to past issues with impropert flagging of channels with 0s, and also the astrophysical occurrence of scintillation in a number of cases.