DM-48591: Add multiprofit two-Gaussian PSF and Sersic columns #309
Conversation
taranu
force-pushed
the
tickets/DM-48591
branch
from
b7af916 to
484ba0a
Compare
| - name: g_psfModel_TwoGaussian_cen_x | ||
| "@id": "#Object.g_psfModel_TwoGaussian_cen_x" | ||
| datatype: double | ||
| description: Centroid (x-axis) of the two-Gaussian PSF model (g-band). |
There was a problem hiding this comment.
This is relative to the declared center of the original PSF model? If so, it might be clearer to call it a delta_x or something.
There was a problem hiding this comment.
No, it's in (multiprofit, not converted to afw/numpy convention) PSF image coordinates, so it's a lot of 17.5 +/- 0.005 values. They're not really useful for anything except as a diagnostic in case it's far off x.5, and I guess you can multiply them by 2 and round to get the PSF stamp size.
| - name: z_psfModel_TwoGaussian_is_parent_flag | ||
| "@id": "#Object.z_psfModel_TwoGaussian_is_parent_flag" | ||
| datatype: boolean | ||
| description: Flag set for objects not fit because they are parents for the two-Gaussian PSF model (z-band). |
There was a problem hiding this comment.
If this is derivable from some other object table flag, I don't think we need to replicate it here, as long as we document that the PSF model Gaussian approximations and galaxy model fits don't run on parent objects.
| description: Flag set for objects not fit because they are parents for the two-Gaussian PSF model (z-band). | ||
| fits:tunit: | ||
| # MultiProFit Sersic model | ||
| - name: sersic_cen_x |
There was a problem hiding this comment.
I would be inclined to drop the pixel-coordinate centers from the standardized tables, since are redundant with the (ra, dec).
There was a problem hiding this comment.
I'm a little hesitant to do this because I derive the ra, dec uncertainties in a fairly stupid way. I just convert x + x_err, y + err to ra, dec and subtract the original ra_dec. This is probably fine in most cases but I don't think I was careful enough with edge cases like RA=360 or dec=-90. I suppose I should use whatever Clare implemented for the ra, dec errors instead.
| datatype: double | ||
| description: Centroid (tract, y-axis) from the multiband Sersic model fit. | ||
| fits:tunit: pixel | ||
| - name: sersic_cen_ra |
There was a problem hiding this comment.
I think cen_ here is redundant; I'd drop it.
| description: Centroid (declination) from the multiband Sersic model fit. | ||
| fits:tunit: deg | ||
| - name: sersic_reff_x | ||
| "@id": "#Object.sersic_reff_x" |
There was a problem hiding this comment.
As I think you're aware, I'm hoping to switch all ellipse-parameter fields to angular coordinates (DM-48425; @ctslater may have another related ticket) and standardize on one ellipse parameterization (probably {a, b, theta} or {r_det, q, theta}?). Since we don't know what we'll switching to yet, I don't think that ought to be part of this ticket, but it's coming.
There was a problem hiding this comment.
Yeah, I haven't polled galaxies about this yet. I'm a bit concerned that transforming the errors is not trivial, even ignoring covariances.
| datatype: double | ||
| description: Sersic profile index parameter from the multiband Sersic model fit. | ||
| fits:tunit: | ||
| - name: sersic_cen_xErr |
There was a problem hiding this comment.
I recommend dropping these pixel-coordinate uncertainties along with the pixel-coordinate values.
| "@id": "#Object.sersic_rhoErr" | ||
| datatype: double | ||
| description: Error on ellipse rho (correlation coefficient) from the multiband Sersic model fit. | ||
| fits:tunit: |
There was a problem hiding this comment.
Do you have any covariance uncertainties for the ellipse parameters + sersic index + flux?
There was a problem hiding this comment.
Not yet, unfortunately.
| datatype: int | ||
| description: Number of iterations in the non-linear fit for the two-Gaussian PSF model (g-band). | ||
| fits:tunit: | ||
| - name: g_psfModel_TwoGaussian_chisq_red |
There was a problem hiding this comment.
I'd just spell out _reduced; _red is too overloaded.
| datatype: boolean | ||
| description: Flag set for failures with an unexpected or unknown cause for the multiband Sersic model. | ||
| fits:tunit: | ||
| - name: sersic_is_parent_flag |
There was a problem hiding this comment.
If this and sersic_not_primary_flag are redundant with the corresponding deblend_ and detect_ flags, I think we can drop them.
taranu
force-pushed
the
tickets/DM-48591
branch
4 times, most recently
from
a55ce7e to
6e1ffda
Compare
taranu
force-pushed
the
tickets/DM-48591
branch
from
6e1ffda to
e73bbd8
Compare
Checklist
When making changes to YAML files in the schemas directory: