Commentary: Astronomical transient detection controlling the false discovery rate

Abstract

The two-step, False Discovery Rate-based thresholding procedures presented by Clements and Sarkar in this volume offer a computationally efficient means by which to detect faint sources lurking in collections of megapixel and gigapixel images. We compare Clements and Sarkar's Procedure 36.1 with the False Cluster Proportion-based algorithm of Friedenberg and Genovese (arXiv:0910.5449, 2009). The former employs pixel-wise error control, while the latter employs cluster-wise error control. We find the two techniques yield source lists of similar efficiency (finding ≈50% of the sources detected by a more computationally intensive procedure) and purity (≈100%), if one eliminates singlepixel detections made by the Clements and Sarkar procedure. We propose that the Clements and Sarkar procedure be refined such that only statistically significant clusters are retained in the final source list, mitigating the issue of single-pixel detections and potentially improving the procedure's efficiency. © Springer Science+Business Media New York 2013.