Eric Bellm: Are you able to review this small-ish ticket? There are two PRs with about 80 lines of changes. Given that this is writing to a dataframe, there may be a faster approach than my getSignificance closure there, but this works fine.

• https://github.com/lsst/meas_base/pull/215
• https://github.com/lsst/ap_association/pull/154

Note that I filed DM-35331 because I realized it would be too much of a pain to add tests of sky sources on this ticket: I'm aware that I'm not testing that path in the significance code, but I did run ap_verify_ci_hits2015, which I think does use that path.

I think there's an indexing bug in ap_association, but otherwise it looks good.

Thank you for catching that bug: I've incorporated your fix and verified that it works on ap_verify cosmos (identical results to main, for the metrics we measure).

Which reminds me: should we have a metric on this detection significance for diaSources? (added on another ticket, probably)

New Jenkins run: https://ci.lsst.codes/blue/organizations/jenkins/stack-os-matrix/detail/stack-os-matrix/36968/pipeline

I don't think it needs a metric, myself--it's already a column in the APDB. What might be more interesting is a metric which compares (in aggregate) the difference between the detection SNR and the measurement SNR. But I think we should work on notebook-level exploration to understand the behavior first.

Thanks John! Looks good now.