Testing multiple-group variance equality with randomization procedures
- 30 August 2007
- journal article
- research article
- Published by Taylor & Francis in Journal of Statistical Computation and Simulation
- Vol. 77 (9) , 797-803
Abstract
The operating characteristics of different statistical tests for evaluating multiple-group variance equality were examined using a simulation study. Normal-theory tests including Bartlett's test, Levene's test and Hartley's test were compared to randomization tests in this study. The randomization tests maintained nominal Type I error rates over a variety of conditions. Levene's test was conservative under all conditions examined. Both Bartlett's test and Hartley's test were anti-conservative (rejects more frequently than nominally specified) when the distribution was lighter-tailed than a normal distribution (double exponential) and were conservative when the distribution was heavier-tailed than a normal distribution (uniform). Bartlett's test performed exceptionally well when the distribution was normal. Use of randomization test is recommended as a test of variance equality as this test performed at nominal Type I error rates and exhibited better power properties than Levene's test, the other test that controlled Type I error rates (albeit conservatively).Keywords
This publication has 9 references indexed in Scilit:
- Resampling Methods2001
- Testing Variances in Psychological and Educational ResearchJournal of Educational Statistics, 1994
- Testing variance equality with randmization testsJournal of Statistical Computation and Simulation, 1989
- Type I Error Rates and Power Estimates of Selected Parametric and Nonparametric Tests of ScaleJournal of Educational Statistics, 1987
- The Role of Permutation Groups in Randomization TestsJournal of Educational Statistics, 1983
- A Comparative Study of Tests for Homogeneity of Variances, with Applications to the Outer Continental Shelf Bidding DataTechnometrics, 1981
- Robust Tests for the Equality of VariancesJournal of the American Statistical Association, 1974
- NON-NORMALITY AND TESTS ON VARIANCESBiometrika, 1953
- Properties of sufficiency and statistical testsProceedings of the Royal Society of London. A. Mathematical and Physical Sciences, 1937