Robustness for Omnibus Tests using Trimmed Means under Violated Assumptions

  • Hyunchul Kim (Department of Education, Sungkyunkwan University, 3-53, Myungryun-Dong, Chongro-Ku, Seoul, Korea, 110-745)
  • Published : 1997.08.01

Abstract

Univariate F test is based on the multisample sphericity assumption. Robustness for tests of a main effect of the within-subjects factor was investigated when the assumptions of the onmibus F tests are violated in a split-plot design with one between-subjects factor using untrimmed data and trimmed data. The results indicate that when sample sizes are unbalanced and dispersion matrices are heterogeneous, the CIGA and the $CIGA_T$ tests better control Type I error rates than do the $F_T$test and the$\widetilde{\xi_T}$test.

Keywords

References

  1. Psychometrika v.24 On methods in the analysis of profile data Greenhouse, S.W.;Geisser, S.
  2. The Korean Communications in Statistics v.4 no.1 Type 1 error rates and power for omnibus tests of repearted measures means in the split plot design Kim, H.
  3. Fundamentals of experimental design(3rd ed.) Myers, J.L.
  4. The American Statistician v.34 no.4 Comparative Analysis of Pretest-posttest research designs Brogan, D.R.;Kutner, M.H.
  5. British Journal of Mathematical and Statistical Psychology v.48 An improved general approximation test for the main effect in a split-plot design Algina, J.;Oshima, T.C.
  6. Exploring data tables,trends, and shapes Summarizing shape numerically: The h-and-h distributins Hoaglin, D.C.;D. Hoaglin(ed.);F. Mosteller(ed.);J. Tuket(ed.)
  7. Multivariate Behavioral Research v.29 Some alternative approximate tests for a split plot design Algina, J.
  8. British Journal of Mathematical and Statistical Psychology v.43 Analysinf unbalanced repearted measure designs Keselman, J.C.;Keselman, H.J.
  9. Sankyha A v.25 Less vulnerable confidence and significance procedures for location based on a single sample: Trimming/Winsorization 1 Tukey, J.W.;Mclaughlin, D.H.
  10. British Journal of Methematical and Statistical Psychology v.32 Analysis of repeated measurements Rogan, J.C.;Keselman, H.J;Mendoza, J.L.
  11. British Journal of MAthematical and Statistical Psychology v.46 Analyzing repeated measures or randomized block designs using trimmed means Wilcox, R.R.
  12. Journal of Educational Statistics v.16 A correction for the ε approximate test in repeated measures designs with two or more independent groups Lecoutre, B.
  13. Annals of Mathematical Statistics v.25 Some theorems on quadratic forms applied in the study of analysis of variance problems: Ⅰ. Effect of inequality of variance on the one-way classification Box, G.E.P.
  14. Psychometrika v.43 no.2 Some approximate tests for repeated measurement designs Huynh, H.
  15. SAS/IML Software:Usage and Reference, Version 6(1st ed.) SAS Institute Inc.
  16. Journal of Educational Statistics v.1 no.1 Estimation of the Box correction for degrees of freedom from sample data in randomized block and split-plot designs Huynh, H.;Feldt, L.S.
  17. Exploratory data analysis Tukey, J.W.
  18. British Journal of Mathematical and Statistical Psychology v.31 Robustness? Bradkey, J.C.
  19. Journal of the Amerucan Statistical Association v.65 Conditions under which mean squared ratios in repeated measurements design have exact F distributions Huynh, H.;Feldt, L.S.