the goal is to keep the overall false rejection probability down at the to level For each validated we discuss which of course is by necessity relative timing We take the time for a single random via one of the fastest good and at any rate the ratios will vary by around depending on the processor and memory configuration used A timing for a distribution of units would mean no time used beyond the uniform random Summary Distribution Validated Validation Rejected Past N RandGauss RandGaussT RandGaussQ bins stepwise bins RandPoisson RandPoissonT mu< 100 N=50, 000, 000 ------- mu > RandPoissonQ mu< 100 N=50, 000, 000 -------(same as RandPoissonT) mu > RandGauss shoot() etc 2.5 units Validation tests applied and a very accurate series is substituted for the table method shoot() etc 1.7 units Validation tests applied and below about **a series approximation is but we have applied this test with N up to and never get anywhere near the rejection point Analytical considerations indicate that the validation test would not reject until O(10 **24) samples were inspected. ---------------------------------------------------------- 2. RandGeneral Method since we wish to have good resolution power even if just one of the mu values is it would be unwise to dial this down any further Validation for several values o mu)