Periodic transit and variability search with simultaneous systematics filtering: Is it worth it?

TL;DR

This study compares traditional and simultaneous filtering methods for detecting periodic signals in photometric data, finding that traditional filtering generally yields higher recovery rates and faster processing, with some nuances for long-period signals.

Contribution

It provides a systematic comparison of filtering methods for signal detection in photometric time series, highlighting the efficiency and limitations of each approach.

Findings

Traditional filtering yields higher signal recovery rates.

Simultaneous filtering performs better with many templates.

Detection rates for long-period signals decrease with more templates.

Abstract

By using subsets of the HATNet and K2 (Kepler two-wheel) Campaign 1 databases, we examine the effectiveness of filtering out systematics from photometric time series while simultaneously searching for periodic signals. We carry out tests to recover simulated sinusoidal and transit signals added to time series with both real and artificial noise. We find that the simple (and more traditional) method that performs correction for systematics first and signal search thereafter, produces higher signal recovery rates on the average, while also being substantially faster than the simultaneous method. Independently of the method of search, once the signal is found, a far less time consuming full-fledged model, incorporating both the signal and systematics, must be employed to recover the correct signal shape. As a by-product of the tests on the K2 data, we find that for longer period sinusoidal signals the detection rate decreases (after an optimum value is reached) as the number of light curves used for systematics filtering increases. The decline of the detection rate is observable in both methods of filtering, albeit the simultaneous method performs better in the regime of relative high template number. We suspect that the observed phenomenon is linked to the increased role of low amplitude intrinsic stellar variability in the space-based data. This assumption is also supported by the substantially higher stability of the detection rates for transit signals against the increase of the template number.