Residuals obtained from least sum of absolute deviations regression are analyzed by a procedure consistent with Euclidean geometry. Applications to a split-plot design are illustrated and evaluated. The procedure is applicable to balanced and unbalanced split-plot designs and avoids the customary assumptions of ordinary least squares analyses including normality, homogeneity of variance and covariance, and sphericity.
Get full access to this article
View all access options for this article.
References
1.
BerryK. J.MielkeP. W.Jr. (1998a) A FORTRAN program for permutation covariate analyses of residuals based on Euclidean distance. Psychological Reports, 82, 371–375.
2.
BerryK. J.MielkeP. W.Jr. (1998b) Least absolute regression residuals: Analyses of block designs. Psychological Reports, 83, 923–929.
3.
BerryK. J.MielkeP. W.Jr. (1999) Least absolute regression residuals: Analyses of randomized designs. Psychological Reports, 84, 947–954.
4.
EdwardsA. L. (1985) Experimental design in psychological research. (5th ed.) New York: Holt, Rinehart & Winston.
5.
KeppelG. (1982) Design and analysis: A researcher's handbook. (2nd ed.) Englewood Cliffs, NJ: Prentice-Hall.
6.
KeppelG.ZedeckS. (1989) Data analysis for research designs: Analysis of variance and multiple regression/correlation approaches. New York: Freeman.
7.
KirkR. E. (1982) Experimental design: Procedures for the behavioral sciences. Belmont, CA: Brooks/Cole.
8.
MielkeP. W.Jr.BerryK. J. (1994) Permutation tests for common locations among samples with unequal variances. Journal of Educational and Behavioral Statistics, 19, 217–236.
9.
MielkeP. W.Jr.BerryK. J. (1997) Permutation covariate analyses of residuals based on Euclidean distance. Psychological Reports, 81, 795–802.
