Define the project

[dkirkby]

Here is a first attempt to define a suitable Sprint Week project based on the simulation files provided by @esheldon. Please add any suggestions or comments below, or define a separate project in a new issue.

The project is to Study the multiplicative biases of the MOF pipeline due to blending as a function of (1) nearest neighbor distance and (2) nearest neighbor flux.

In more detail, p.14 of @esheldon's Nov. 6 talk reports ~2% multiplicative biases using übserseg only, but does not show how biases depend on the properties of individual blends.

The type of plots we are aiming for could be compared with these ones from Samuroff++ 2017 (see also Simon Samuroff's talk from previous week):

[esheldon]

That would be an interesting test.

Note that the selection effects should be calculated in this scenario, using the metacalibration formalism. This should be done separately for each such bin in distance or mag. Selection effects can in general be at the few percent level, so could easily be comparable to neighbor bias.

[esheldon]

Here is a first go at this. This is a only subset of all the data.

As with all correlation functions, the error bars are correlated

Note the x axis is somewhat mislabled, it is binned in distance to nearest neighbor in arcsec. To compare with Simon's plot which is in pixels, the pixel scale is 0.263 for both this sim and his

[esheldon]

Also that is s/n > 10 and T/Tpsf > 0.5 and some flag cuts

[esheldon]

Now compare to the version with the ubserseg masking only, no deblending