[Previous] | [Session 30] | [Next]
C.J. Miller, R.C. Nichol, L. Wasserman, C. Genovese (CMU), A. Connolly, A. Hopkins (Univ. of Pittsburgh)
Source detection in astronomical images is a challenge across wavebands. Common to all source detection algorithms, is determining whether a pixel is part of the background or part of a source. No matter what the threshold, we will always incorrectly classify some pixels. It is common practice to choose a significance threshold that attempts to maximize the number of true sources detected, while minimizing the number of background pixels erroneously identified as sources. However, any such threshold is arbitrary, with some researchers accepting 3 sigma detections, while others demanding 7 sigma detections. We discuss how the meaning of any arbitrarily chosen threshold depends on the intensity distribution of the sources and the background, as well as the true number of source and background pixels. Since one can never know these measures a priori, any arbitrary threshold is inherently meaningless. However, there is one new thresholding technique, entitled False Discovery Rate (FDR), which allows the researcher to specify the maximum number of mistaken background sources. Additionally, FDR provides a meaningful statistic that does not vary with the background or source distributions. FDR is easy to implement and can be tuned to the application at hand. We show some examples using radio data, as well as optical data from the Hubble Deep Field.
If you would like more information about this abstract, please follow the link to http://www.cjmiller.net/research. This link was provided by the author. When you follow it, you will leave the Web site for this meeting; to return, you should use the Back comand on your browser.
The author(s) of this abstract have provided an email address for comments about the abstract: chrism@cmu.edu