200 likes | 363 Views
STEP4. Konrad Kuijken Leiden. Back to basics?. STEP1,2,3 ‘end-to-end’ tests for constant-shear images What causes discrepancies? Selection bias PSF modelling error Ellipticity measurement error function of (mag, size) Noise biases Neighbours Simulation errors? … …
E N D
STEP4 Konrad Kuijken Leiden
Back to basics? • STEP1,2,3 • ‘end-to-end’ tests for constant-shear images • What causes discrepancies? • Selection bias • PSF modelling error • Ellipticity measurement error function of (mag, size) • Noise biases • Neighbours • Simulation errors? • … … • Controlled simulations of effects separately
STEP4 • Concentrate on ellipticity and PSF: • Reveal positions to within 1 pixel • Well-separated sources • Input catalogues rigorously with <ei(e)>=0 • Brute-force PSF convolution, pixellation: sample • Different S/N simulations • Build up complexity gradually (gal, PSF) • Blind catalogues, some shear values public
STEP4 Images Galaxy models: Exp & deVauc fwhm 7pix Moffat PSF fwhm 3.5pix • 3720x3720 pixel fits files • 60x60 grid of galaxies • 60-pixel separation • 240 stars around the sides • Stars 10x as bright as galaxies • 32 different shear values, ePSF per simulation set ()~10-4
Ellipticity distributions • From Lambas et al 1992 (APM survey)
How to analyze STEP4 data? • First 8 shear & ePSF values are public (2%) • If your method can reproduce these with m<1%, then send me all results to get a measurement of m,c • Intended as a resource for method development - prerequisite to test other nuisance effects
STEP4 - next simulations? • PSF, galaxy sizes • Galaxy type (varying ellipticity; spiral arms; irregulars) • PSF type • About 30GB per set of simulations (32 x [10+5+2] images)
Summary • 100,000,000 simulated galaxies available • S/N=15,50,200 • Galaxies with sersic n=1,4, fwhm 7pix • PSF Moffat, FWHM 3.5pix • Well-controlled images • No shape noise • Statistical noise ()~10-4 per gal/PSF/SNR simulation • More simulations possible (just need a MC sample of 107 ‘photons’ from galaxy and PSF