Inference for negativist theory using numerically computed rejection regions

A negativist theory is a theory that asserts that an a0ect is absent. Negativist theories are common in statistics and require a test procedure to avoid biasing the conclusions in favor of accepting the theory. New critical value tables are provided and computational methods are introduced to ease practical di3culties with the resulting non-contiguous interval null. Simulations compare the power function of negativist testing to traditional hypothesis testing. A case study, using Canadian stock market data, does not always support e3cient markets (an example of a negativist theory) and reverses previous studies using positivist hypothesis tests. c © 2002 Elsevier Science B.V. All rights reserved.