Open stscijgbot-jp opened 2 years ago
Comment by Howard Bushouse on JIRA:
Now that #6739 has been merged, which resulted in closing JP-2284, can this ticket also be closed? Is there any other work left to do for it?
Comment by Jane Morrison on JIRA:
No I am working on the last part of this - step 5 (above) removing the evidence function use from the Bayesic package and using astropy
Jane Morrison Remind me where we are with this? Things that you're actively working on still or is this one of many potential long-term tweaks?
Issue JP-2344 was created on JIRA by Jane Morrison:
JP-2284 and #6739 (replacement for #6475) addresses the first step to removing the dependence on the 3rd party package - BayesicFitting. For this ticket a comparison of the two methods will be done with an emphasis on evaluating with method removes the fringes better.
Update: received new information and updated algorithms from Patrick Kavanagh on 12/22/21. Patrick has updated the method routine to derive the fringes (update to fit_1d_fringes_bayes_evdience). This new method uses new CDP files. Patrick also supplied some test data where the number of fringes and their frequencies and amplitudes are known. We can use these in a test. They are three files with " The header of each file is a dict which contains the number of fringes included and their parameters. The parameters in order are freq_1, cos_amp_1, sin_amp_1, freq_2, cos_amp_2, sin_amp_2, ..., freq_n, cos_amp_n, sin_amp_n. Note that the frequencies are straight from the best fit model and have not been multiplied by the scaling factor as is done at the end of fit_1d_fringes_bayes_evidence. The columns are wavenumber, flux, weights.
This ticket may be broken into separate tickets. The works that needs to be done includes: [0.5] 1. Add unit tests - (could we use the data Patrick has made in a unit test) A unit test for the background was added as part of #6739. . Need to add a unit test to when evidence function is part of astropy
[X] 2. Add regression tests - added in #6771
[X] 3. Test JP-2284 to confirm astropy routines for fit_1d_background_complex work as well as Bayesic Fitting routines (#6739) [X] 4. Replace the fit_1d_fringes_bayes_evidence function with new routine from Patrick
[] 5. Compare the second part of removing the dependence of the BayesFitting package which involves replacing the routines in fit_1d_fringes_bayes_evidence with astropy routines