search

annayqho / TheCannon

a data-driven method for determining stellar parameters and abundances from stellar spectra
MIT License
39 stars 16 forks source link

weird alpha plume #61

Closed annayqho closed 8 years ago

annayqho commented 9 years ago

things to try: -- train only on the good stars, test on good+bad -- train on bad, test on bad -- train on partial regions of the spectrum to see if there are particular regions causing the model to break

davidwhogg commented 9 years ago

can you show me a figure?

mkness commented 9 years ago

The [Ca/Fe] and [Mg/Fe] versus alpha/Fe relations show a large scatter (partially there is a larger offset at lower SNR but also high SNR are offset) > Are the stars that are up in the offset alpha-plume in LAMOST also the ones that are the furthest from the 1:1 relation in the Ca/Fe v. alpha/Fe results from APOGEE (or Mg/Fe v. alpha/Fe) ? I'm wondering what the effect would be if elements dominating the overall alpha/Fe result from APOGEE are dissimilar to the elements that dominate in LAMOST - i.e a potential issue if the scale between individual alpha elements is offset. Had not considered this. Also, the stars that are in the offset alpha-plume do not correspond to the low SNR Apogee stars do they?

davidwhogg commented 9 years ago

I remain concerned that this could be continuum. I want to see exactly what's being done there.

annayqho commented 8 years ago

convergence failure! yay!