The direct need is to confirm that the MCS centroids are not being biased by the centroiding algorithm being used. One way to do this would be to estimate an unweighted mean position, or to use a different Gaussian (e.g. twice as wide). Another would be to use a PSF estimating code such as piff, and compare the unweighted centroid of the PSF model with the results of applying the MCS algorithm. While doing this we'll have to be careful about noise-weighting; the simplest thing to do is to ignore the image variance entirely (although it is not entirely clear that we can afford to do this with the real MCS data)

Report on this work already filed.