Question: how to parameterize polynomial baseline functions?

[jpdeleon]

I searched but couldn't find in examples and documentation a params.csv file that uses hybrid_poly_2 (and higher order ones) baseline function. I tried to generate the params.csv file using the GUI but when, for example, hybrid_poly_2 is selected, no relevant parameter appears in the generated params.csv. Actually, why no parameter is specified in params.csv when using any hybrid_* baseline functions?

Anyway, I proceeded to run allesfitter.show_initial_guess but it appears from the resulting plots that the baseline function is constant despite changing the order of hybrid_poly_* baseline function in the settings.csv file. hybrid_spline however works fine.

[MNGuenther]

Hi @jpdeleon, thanks a lot for raising this -- and apologies that this fell through the cracks so far.

Just to explaind: All baseline options with hybrid_* will not show up with parameters anywhere in allesfitter. That is because they are fitted using leastsquares optimization directly with scipy (and other) modules. For example, the hybrid_offset will simply subtract the median of the residuals. The hybrid_spline and hybrid_poly will be optimised by scipy itself. It is super rare that this "fails" and I have only seen it once so far: in the presence of extremely rapid systematic noise such as pulsations of <1 hour periods. In that case, scipy gets confused and sees the whole pattern as white noise. That is why it then draws a straight line through it all instead of trying to take it out.

I will close this for now, as it is highly case specific. If this still persists or reoccured for you, please send along a Dropbox/Google Drive link to your folder via email and I will look into it!