Multivariate profile analysis is proposed as an alternative to Ghiselli's correlation method of identifying moderator variables for the following reasons: (a) The direction of predictability is not sacrificed, since absolute values of residuals are not used, (b) A multivariate method of analysis is needed when screening more than one potential moderator, (c) Residual profiles provide more information than single correlation coefficients.
References
1.
BrownF. G.ScottD. A.Differential predictability in college admissions testing. Journal of Educational Measurement, 1967, 4, 163–166.
2.
CarverR. P.The case against statistical significance testing. Harvard Educational Review, 1978, 48, 378–399.
3.
GhiselliE. E.Differentiation of tests in terms of the accuracy with which they predict for a given individual. Educational and Psychological Measurement, 1960, 20, 675–684.
4.
GhiselliE. E.Moderating effects and differential reliability and validity. Journal of Applied Psychology, 1963, 47, 81–86.
5.
MorrisonD. F.Multivariate statistical methods. New York: McGraw-Hill, 1976.
6.
SaundersD. R.Moderator variables in prediction. Educational and Psychological Measurement, 1956, 16, 209–222.
7.
TuckmanB. W.Conducting educational research. New York: Harcourt Brace Jovanovich, 1972.
