Statistically Proper Multiple Range Tests for a Within Subject Factor in a Repeated Measures Design

다국어 초록 (Multilingual Abstract)

It is a common practice in many research areas that multiple range tests for a between subject factor such as Tukey are applied to a within subject factor in a repeated measures design. Tukey procedure, however, sometimes detects no pairs with different means even when the hypothesis of all equal level means is rejected. This study attempts to provide a rationale for the proposition that Tukey is inappropriate post hoc procedure for a within subject factor in which the observations are correlated. We introduce two multiple range tests, Bonferroni and Scheffe, for a within subject factor and show that Bonferroni is more appropriate than Scheffe for pairwise multiple comparisons. Subsequent simulation study indicates that Tukey has significantly less power than Bonferroni in detecting actual difference between means of some pairs when the observations of a within subject factor are highly correlated.

참고문헌 (Reference)

1 Teixeira, L. A., "Timing and force components in bilateral transfer of learning" 44 : 455-469, 2000

2 Barthelemy, S., "Orienting visuospatial attention generates manual reaction time asymmetries in target detection and pointing" 133 : 109-116, 2002

3 Harvey, M., "Manual responses and saccades in chronic and recovered hemispatial neglect : a study using visual search" 40 : 705-717, 2002

4 Helsen, W. F., "Coupling of eye, finger, elbow, and shoulder movements during manual aiming" 32 : 241-248, 2000

5 Benjamini, Y., "Controlling the false discovery rate : a practical and powerful approach to multiple testing" 57 : 289-300, 1995

6 Johnson, R. A., "Applied Multivariate Statistical Analysis, third ed" Prentice Hall 1992

7 Neter, J., "Applied Linear Statistical Model, second ed" Irwin 1985

8 Anderson, T. W., "An Introduction to Multivariate Statistical Analysis, second ed." John Wiley & Sons, Inc. 1984

9 Weir, P. L., "Age-related differences in prehension : the influence of task goals" 30 : 79-89, 1998

10 Benjamini, Y., "A distribution-free multiple test procedure that controls the false discovery rate" Department of Statistics and O.R., Tel Aviv University 1999