Response to Comments on Draft
Wednesday 7th November, 2012
-
Goals
-
Statistical framework Calibration to kappah, why this fails (Figure 9) Calibration via kappah_median, how this relates to Zach's work Updated Figures following today's visit: these now match the text, see below.
-
Investigation of selection bias As an example of a source of systematic error Need to present carefully (Tommaso, Sherry) - getting balance right between investigating the current (and next gen) situation, and exploring what needs to be done for surveys beyond that.
-
Weak lensing limit Required by sims we have, mag equations valid (but could be better motivated, Sherry) Means we are limited to estimating kappa for line of sight, not local environment - needs spelling out Relevance to Suyu et al, Green et al: also using weak lensing kappa from sims, but N45 that includes local environment.
-
Conclusions about spectroscopic data?
Long-term:
-
To implement a flexible halo model analysis framework capable of bringing all available observational data to bear on the problem of estimating the weak lensing convergence or magnification at any point in the sky.
-
To test this analysis for systematic errors with realistic simulated data.
Short-term: moving beyond Suyu et al 2010, Green et al 2012.
-
To investigate the simplest possible halo model, using just "observed" stellar masses and redshifts (of varying precision) for each galaxy, quantifying the expected precision and accuracy of the reconstructions.
-
To include and test simple prescriptions for accounting for voids and unseen halos, based on calibrating the halo model predicted convergence to the ray-traced "true" convergence.
-
As one example of an investigation of a source of systematic error: to what extent does the reconstruction method preserve its accuracy for different sightline selections. Random should work by construction; other selections may not. Given this very simple reconstruction method, and some plausible choices of subsample, what systematic errors might creep in?
Calibrating via Pr(kappah|data) from calibration lines of sight in simulations. Fails when we only have photoz, due to high mass tail.
Instead, use median kappah as indicator, and look up Pr(kappa|kappah-median). This is now fixed, to match text. Precision, accuracy shown below. Downside is that this method is more strongly tied to sims (in that many calibration reconstructions are required).
Median kappah is equivalent quantity to weighted N45 - as in Suyu et al 2010, 2012, Greene et al 2012.
