bd-j
commented
3 years ago A few things that could be important
1) Do the fits look reasonable, or is the photometry being massively misfit when including the spectrum?
- Is the spectral resolution (physical & instrumental) being fit? It looks like not. See e.g.
TemplateLibrary["spectral_smoothing"]- Are the spectra somehow normalized to the photometry? If not are you including some spectroscopic calibration model? (It looks like the 2nd answer is no, you'd want to use
PolySpecModeland addTemplateLibrary["optimize_speccal"])
- Are the spectra somehow normalized to the photometry? If not are you including some spectroscopic calibration model? (It looks like the 2nd answer is no, you'd want to use
2) The spectrum might actually be informative about the age or metallicity in a way that shifts the posterior from photometry only. You might compare the posterior distributions for logzsol and the SFH (also for dust)
CSwigg
This may be a more general question/issue about SED fitting, but I am finding some discrepancies in the output stellar mass (surviving) when running prospector on a sample of 50 galaxies using just their photometry (UV through mid-IR) and then again for a run including their optical spectra. My model is non-parametric with a continuity prior; here is a print out:
I am finding that there is a ~0.25 dex shift in stellar mass between the two runs. I am beginning to consider some of the 'Fitting Spectra' bullet points in the Prospector FAQ, but are there any other more obvious reasons why including the spectra might create this shift? Here I am just including the spectra in a 3600-4200 AA region around the 4000 AA break, but using the full spectra also creates a similar shift. Let me know if I can provide any more info to make my procedure more clear. Also tagging @moustakas on this. Thank you!