Testing for impulsive behavior: A bootstrap approach

Increasingly, systems are being designed to account for impulsive behavior that may be present in signals, with one of the widely used statistical models used being the alpha stable distribution. Two techniques are presented that test for the level of impulsive behavior: testing the parameter alpha directly and a characteristic function (cf) based technique. The parametric bootstrap is used in both cases to estimate the distribution of the test statistics and in the setting of critical values. Simulation results show both tests maintain their level and achieve high rejection rates under alternatives. (C) 2001 Academic Press.