[Previous] | [Session 54] | [Next]
A. Connors (Eureka Scientific), D.A. Smith, M. Muno (MIT CSR)
An early serendipitous detection{*} of a variable X-ray afterglow in the hour after a burst gave impetus to tracking down similar phenomena in a more systematic way with other instruments such as RXTE's All Sky Monitor (ASM). However the signal--to--noise ratio could be quite low. Bayesian methods always allow one to derive the best possible measure of any model or hypothesis, given the data --- although the result is not always fast. Up until now, the RXTE ASM relied on a very fast, robust, direct matrix inversion technique, with a \chi2 approximation. We are now in the process of implementing additional Bayesian methods. We start with the full Poisson likelihood, then marginalize over any uninteresting parameters, to get more specifically designed (if slower) likelihood ratios and credible regions for new sources. These methods should give similar results when the measured source counts are many of sigma above the background; but for fainter sources we expect more accurate constraints. We will describe our method, and results to date. This is funded by NASA Grant NAG5-8476.
{*} (In HEAO 1 A2 Data; see Connors and McConnell 1995 Rome ICRC, 2, 57; Connors and Hueter 1998 ApJ 1998ApJ 501, 307 and references therein.)