STEP: Cosmic shear analysis of image simulations
STEP home Simulations Data Mailing List

The STEP image simulations have undergone a blind cosmic shear analysis by many different researchers using many different shear measurement techniques. Here we present a summary report which contains all our key findings, and in addition we provide supplementary material for the interested reader.

Please e-mail comments, queries and suggestions to the STEP mailing list, STEP[at]mpia.de.

Shear TEsting Programme: Cosmic shear analysis of image simulations

Catherine Heymans, Ludovic Van Waerbeke, Richard Massey, David Bacon, Gary Bernstein, Emmanuel Bertin, Sarah Bridle, Michael L. Brown, Douglas Clowe, Håkon Dahle, Thomas Erben, Meghan Gray, Marco Hetterscheidt, Henk Hoekstra, Mike Jarvis, Konrad Kuijken, Vera Margoniner, Yannick Mellier, Reiko Nakajima, Molly Peeples, Alexandre Refregier, Jason Rhodes, Tim Schrabback, David Wittman.

Abstract
The Shear TEsting Programme, STEP, can improve the accuracy and reliability of all weak lensing measurements through the rigorous testing of cosmic shear pipelines, the exchange of data and the sharing of technical knowledge. In this report we present the STEP analysis of sheared image simulations, comparing the results obtained from different shear measurement methods. For all but two of the authors this shear measurement was performed without knowledge of the input simulation shear. The most successful method over a wide range of shear distortions 0.0 < &gamma < 0.1 was found to be the Hoekstra implementation of KSB, which produces results that are accurate to better than the 2% level. Similarly accurate results are found in the weak &gamma<0.01 regime with the Bernstein and Jarvis technique. Overall there is an interesting scatter in the results that reveals both residual PSF distortions and calibration bias in some, but not all, methods. In most cases PSF systematics are successfully below the 0.1% level, although the accuracy of the PSF correction is found to be strongly dependent on the form of the PSF distortion. Calibration errors from real data cannot be detected through diagnostic tests such as an E/B mode decomposition and so our finding that shear calibration is often only accurate to ~10% (or worse) in ~50% of the methods tested, causes some concern. Different implementations of the KSB formalism produces remarkably different results and so we provide details of different KSB cosmic shear pipelines to aid the improvement of future KSB analyses. Alternative methods to KSB and Bernstein & Jarvis, appear to fare only as well (or worse) when compared to the average results of the commonly used KSB method. We determine the accuracy of source detection, measure centroid errors and determine levels of selection bias and weight bias. We also investigate the dependence of shear measurements on galaxy properties such as magnitude and size. For some methods we can conclude that the measured calibration bias is related to selection bias and/or the use of biased weights. For other methods the calibration biases appear to arise from the measurement method itself.

The report can be accessed on-line here or downloaded in a postscript file STEP_report.ps.gz .

Supplementary material that is summarised in the report, can be viewed in postscript form;
KSB_analysis.ps ,
other_analysis.ps,
shapelets_analysis.ps,
shear_from_real_false_detections.ps ,
mag_bin.ps,
rg_bin.ps,
disk_bin.ps,

If you wish to perform your own comparison analysis you can download participants catalogues from ftp.mpia.de. The pub/COMBO/STEP_sim_cats directory contains the following tar directories:
SkyMaker simulation catalogues SkyMaker.tar.gz
Shapelet simulations catalogues shapelets.tar.gz
Input SkyMaker catalogues input_cats.tar.gz
where the format of the input catalogues are described here.

The minutes of the December STEP telecon can be found here


Last modified 3rd December.
STEP pages maintained by Catherine Heymans: heymans[at]mpia.de