Hi @moustakas
I compared the Fastspecfit measurements with those from another spectral-decomposition algorithm for QSOs in the GQP-Y3 test catalog (https://github.com/cosmodesi/desi-y3-gqp/blob/main/notebooks/intro.ipynb). From this catalog, I selected a subsample of 20735 QSOs with redshift 0.59<z<2.22 and r-band magnitude r<21.5. The first figure shows the broad Mg II width from Fastspecfit as a function of redshift, while the second figure shows the corresponding widths obtained by the IFit algorithm developed by Yu et al. 2021, MNRAS, 507, 3771. The third figure plots the difference between the two measurements against redshift. Overall, the Fastspecfit line widths are systematically larger and exhibit greater scatter than those from IFit. Moreover, the Fastspecfit widths exhibit redshift-dependent features that are not seen in the IFit measurements, such as a break of distribution at z ~ 1.6.
Hi @moustakas
I compared the Fastspecfit measurements with those from another spectral-decomposition algorithm for QSOs in the GQP-Y3 test catalog (https://github.com/cosmodesi/desi-y3-gqp/blob/main/notebooks/intro.ipynb). From this catalog, I selected a subsample of 20735 QSOs with redshift 0.59<z<2.22 and r-band magnitude r<21.5. The first figure shows the broad Mg II width from Fastspecfit as a function of redshift, while the second figure shows the corresponding widths obtained by the IFit algorithm developed by Yu et al. 2021, MNRAS, 507, 3771. The third figure plots the difference between the two measurements against redshift. Overall, the Fastspecfit line widths are systematically larger and exhibit greater scatter than those from IFit. Moreover, the Fastspecfit widths exhibit redshift-dependent features that are not seen in the IFit measurements, such as a break of distribution at z ~ 1.6.