We discuss the STEM (Smoothing and Testing Multiple hypotheses) procedure to search for point sources in Cosmic Microwave background maps; in particular, we aim at controlling the so-called False Discovery Rate, which is defined as the expected value of false discoveries among pixels which are labeled as contaminated by point sources. STEM is based on the following four steps: (1) needlet filtering of the observed CMB maps, to improve the signal to noise ratio; (2) selection of candidate peaks, i.e., the local maxima of filtered maps; (3) computation of p-values for local maxima; (4) implementation of the multiple testing procedure, by means of the so-called Benjamini–Hochberg method. These procedures are also implemented on the latest release of Planck CMB maps.

我们讨论了STEM（平滑和测试多个假设）过程，以在宇宙微波背景图中搜索点源；特别地，我们旨在控制所谓的误发现率，该误率定义为标记为被点源污染的像素之间的误发现的期望值。STEM基于以下四个步骤：（1）对观察到的CMB图进行针刺滤波，以提高信噪比；（2）选择候选峰，即过滤后的图的局部最大值；（3）计算局部最大值的p-值；（4）通过所谓的Benjamini-Hochberg方法实施多重测试程序。这些过程也在最新版的Planck CMB地图上实现。