Variability Indices

I was wondering how gPSFluxStetsonJ > 20 was chosen as a filter in @jeffcarlin 's notebook tutorial 7b. I’m reading about Stetson J indices, but I haven’t found any information about the expected values or ranges for, say, RR Lyrae stars.

Hi Bob – That choice was somewhat arbitrary, because I didn’t readily find a suggested range from the literature. What I did was try values to see what did a good job of picking out the true RR Lyrae (since we have the truth tables). Curious to hear what you find out!

Where can I find out how the [u, g, r, i, z, y] PSFluxStetsonJ columns in the DP0.2 DiaObject table are calculated?

Hi Bob – The documentation for the task doesn’t offer much detail, but you can examine the code in the
Python class to see how it is done. In particular, the _stetson_J method has the detailed calculation.

Thank you, @jeffcarlin, for tracking down a link to the code. The _stetson_J method cites Stetson’s paper for what is computed: On the Automatic Determination of Light-Curve Parameters for Cepheid Variables - Astrophysics Data System. The algorithm isn’t quite trivial. Stetson starts with a fairly straightforward excess variance calculation, but it doesn’t work well in practice. So Stetson “robustifies” it by a somewhat ad hoc modification of the weighted mean used to define variability residuals, devised to downweight outliers. A tuning parameter in the robustified mean controls the downweighting, and gets set via an iterative algorithm. There’s not much in the paper in terms of a formal (e.g., model-based) justification of the robustified mean, or study of its statistical behavior. Seems like there is an opportunity here…

Hi,

You might want to try contacting Peter Stetson directly: peter.stetson@nrc-cnrc.gc.ca. He’s usually pretty responsive.

SG

Thank you @TomLoredo and @jeffcarlin! The code is just what I was looking for, Jeff, and the explanation helps, Tom, because I’m still trying to understand the weighting factors. I’m reading Stetson 1996 and some other papers, so it’s nice to see how it’s used. I’m also wondering about adding skewness and kurtosis parameters.

Good idea, @StephenGwyn, I’ll do that!

Interesting discussion, all! I’ll mark one of @jeffcarlin’s replies as the solution.