Details
-
Type:
Story
-
Status: Done
-
Resolution: Done
-
Fix Version/s: None
-
Component/s: pipe_analysis
-
Labels:None
-
Story Points:5
-
Epic Link:
-
Sprint:DRP S17-2, DRP S17-3, DRP S17-4, DRP S17-5, DRP S17-6
-
Team:Data Release Production
Description
Include ellipticity residual comparison and quiver plots in the analysis script, including:
Intra-stack:
scatter/histogram/sky plots for psfUsed stars of
> residual e1 & e2 ellipticities between the source and model psf at the position of the source (for both SDSS and HSM measurements), where:
e1 = (shape_xx - shape_yy)/(shape_xx + shape_yy)
|
e2 = 2*shape_xy/(shape_xx + shape_yy)
|
> residual ellipticity, δe, between source and model psf at source position quiver plot where:
δe = sqrt((e1_src - e1_psf)^2 + (e2_src - e2_psf)^2))
|
Inter-stack:
scatter/histogram/sky comparison plots for;
> comparison of the trace radii of the sf models at the position of psfUsed stars matched between the two stack catalogs, where:
traceRadius = sqrt(0.5*(shape_xx + shape_yy))
|
Attachments
Issue Links
Activity
The following show the plots for a visit level (processCcd + meas_mosaic) run:
and for a tract in the COSMOS HSC RC dataset:
The function added looks like:
def getQuiver(x, y, e1, e2, ax, color=None, scale=3, width=0.005, label=''):
|
"""Return the quiver object for the given input parameters"""
|
theta = [np.math.atan2(a, b)/2.0 for a, b in zip(e1, e2)]
|
e = np.sqrt(e1**2 +e2**2)
|
c1 = e*np.cos(theta)
|
c2 = e*np.sin(theta)
|
if color is None:
|
color = e
|
q = ax.quiver(x, y, c1, c2, color=color, angles='uv', scale=scale,
|
units='width', pivot='middle', width=width, headwidth=0.0,
|
headlength=0.0, headaxislength=0.0, label=label)
|
|
|
return q
|
But for the above plots I used scale=2 and width=0,002.
Bob Armstrong, would you mind giving these plots a quick glance to ensure they look right? There's no need to review any code, but you can have a look if you'd like on the u/lauren/working branch in pipe_analysis on lsst-dm github.
Bob Armstrong, Lauren MacArthur — this has been "in review" for a couple of months now. Any chance we can close it out? I'd love to shut it down for the end of the S17 cycle.
This latter plot that revealed a systematic 0.5% difference between the model psf trace radii
between the HSC-1382 RC dataset run and the LSST run of
DM-6816(I do realize I will have to fix the binning in the right histogram!):