Accuracy of likelihood doesn't increase with the number of bins

Kruthi24 commented 3 years ago

When we change the number of bins by changing epsilon, the accuracy doesn't seem to improve. To do:

[x] Make setup_bins more readable. (@adivijaykumar #16)
[x] Crosscheck the likelihood comparison plots for the bilby case. (@adivijaykumar)
[x] Compare the setup_bins function in bilby to the one in IAS code. (@Kruthi24)
[x] Incorporate masking in get_detector_response. (@Kruthi24)
[x] Check if inteferometer.frequency_array and waveform_generator.frequency_array are consistent. (@Kruthi24)
[x] Check for errors inlal_binary_black_hole_relativebinning and lal_binary_neutron_star_relativebinning. (@Kruthi24)

Kruthi24 commented 3 years ago

The frequencies at the bin edges obtained using setup_bins in bilby and IAS seem to match. In fact, the bilby one is better because the implementation in IAS code seems to produce duplicate values for 1 or 2 bin edges.

Kruthi24 commented 3 years ago

inteferometer.frequency_array and waveform_generator.frequency_array match!

adivijaykumar commented 3 years ago

Closed with #19

adivijaykumar / bilby

Accuracy of likelihood doesn't increase with the number of bins #15