Community feedback welcomed on potential timeseries features

DMTN-118 is a new technote that discusses Rubin’s plans for the timeseries features it will compute on alert and Data Release lightcurves. Timeseries features touch a wide variety of scientific and technical issues, so we want to ensure that we hear from the science community before we make any final decisions.

The current draft of DMTN-118 provides a literature review and discusses some open questions. We are maintaining a working draft feature list at https://ls.st/fkr. Feedback on both documents is welcome.

We plan to present this material in a variety of community forums (SCMA VII, the PST/Science Collaboration Chairs talk series, at the Project and Community Workshop), so there will be many opportunities to comment. However, we also welcome asynchronous discussion of the document here on Community, either in reply to this post or as new topics.

2 Likes

Hi Eric, the review is looking great. Can I check what the thinking is regarding periodicity searches accommodating various noise processes/shapes (I couldn’t see it on my quick read but perhaps I missed it)? Specifically, Vaughan et al. (2016) looked at how red noise tends to produce periodic looking behaviour for a few cycles on long timescales and phase folding would then likely produce a decent chi^2 (this could then produce a load of spurious signals).

Indeed, false periodicities due to correlated noise are a major concern. I’m aware of the Delisle et al. 2020 work extending False Alarm Probabilities to consider correlated noise–other pointers are welcome!