Robust I-Sample Analysis of Means Type Randomization Tests for Variances

摘要

The advent of powerful computers has brought about the randomization technique for testing statistical hypotheses. Randomization tests are based on shuffles or rearrangements of the (combined) sample. Putting each of the I samples \u22in a bowl\u22 forms the combined sample. Drawing samples \u22from the bowl\u22 forms a shuffle. Shuffles can be made with or without replacement.In this thesis, analysis of means type randomization tests will be presented to solve the homogeneity of variance problem. An advantage of these tests is that they allow the user to graphically present the results via a decision chart similar to a Shewhart control chart. The focus is on finding tests that are robust to departures from normality. The proposed tests will be compared against commonly used nonrandomization tests. The type I error stability across several nonnormal distributions and the power of each test will be studied via Monte Carlo simulation.