CMStatistics 2017: Start Registration
View Submission - CFE
Title: Local alternatives of signal detection tests Authors:  Klaus Poetzelberger - WU Vienna (Austria) [presenting]
Abstract: A principal components analysis of local alternatives of certain one-sided tests for the signal detection problem is presented. These tests reject a null hypothesis if a certain diffusion process crosses a specified boundary $b$. Identifying local alternatives with high power is thus equivalent to identify directions from $b$ with maximal increasing boundary crossing probabilities (BCP). In cases, where the BCP is given in closed form, the alternatives can be computed analytically. In all other cases only numerical solutions are possible. We discuss how local alternatives with high power depend on the choice of $b$.