Randomization and Alternative Tests.
Abstract
General randomization test procedures and their applicability as practical tests of significance are discussed. Specific procedures are detailed for the two sample comparison of means and the one-way analysis of variance. Through Monte Carlo simulation, the robustness and power of these specific randomization tests are examined and compared against parametric, nonparametric, and approximate randomization test alternatives. Selected test conditions include various sample sizes, continuous and discrete sampling distributions, and various approximate randomization test sample sizes. Results of the simulation indicate that randomization and approximate randomization tests are as robust and powerful as parametric tests and more robust and powerful than comparable nonparametric tests. Furthermore, the results imply that parametric and approximate randomization tests may provide excellent alternatives to randomization tests when exact randomization tests may be infeasible. (Theses) (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1986
- Accession Number
- ADA178867
Entities
People
- Christopher C. Whitehead
Organizations
- Naval Postgraduate School