Analysis of Covariance: A Proposed Algorithm

Abstract

Many researchers erroneously think that analysis of covariance (ANCOVA) should be performed only when there are pre-existing significant differences between the groups on the potential covariate. It has been shown however that, when correctly applied, ANCOVA offers two major advantages over simple analysis of variance (ANOVA): (a) greater statistical power due to a reduction in error variance; and (b) a reduction in bias in experiments where differences between groups exist at the beginning of an experiment. An algorithm has been proposed where the use of ANCOVA is conditional upon (a) a test of homogeneity of within-group regression slopes; (b) a Pearson correlation coefficient 2 0.3 between the covariate (X) and the dependent variable (1) in the case of a randomized study; in a non-randomized study (e.g. intact groups), the use of ANCOVA is not conditional on r5? 0.3 because, in this case, critical adjustments can be obtained with correlation lower than 0.3; and (c) a linear relationship between X and Y

Get full access to this article

View all access options for this article.

References

Adams, K. M. , Brown, G. G. , and Grant, I. (1985). Analysis of covariance as a remedy for demographic mismatch of research subject groups: Some sobering simulations. Journal of Clinical and Experimental Neuropsychology, 7, 445-462.

Bonett, D. G. (1982). On post-hoc blocking. Educational and Psychological Measurement, 42, 35-39.

Cox, D. R. (1957). The use of a concomitant variable in selecting an experimental design. Biometrika, 44, 150-158.

Feldt, L. S. (1958). A comparison of the precision of three experimental designs employing a concomitant variable. Psychometrika, 23, 335-353.

Fisher, R. A. (1948). Statistical methods for research workers (10th ed.). New York: Hafner.

Huitema, B. E. (1980). The analysis of covariance and its alternatives. New York: Wiley.

Jennings, E. (1988). Analysis of covariance with nonparallel regression lines. The Journal of Experimental Education, 56, 129-134.

Johnson, P. 0. and Fay, L. C. (1950). The Johnson-Neyman technique, its theory and application. Psychometrika, 15, 349-367.

Keppel, G. (1991). Design and analysis: A researcher's handbook (3e ed.). Englewood Cliffs, NJ: Prentice-Hall.

10.

Kerlinger, F. N. and Pedhazur, E. J. (1973). Multiple regression in behavioral research. New York: Holt, Rinehart and Winston.

11.

Kirk, R. E. (1982). Experimental design: Procedure for the behavioral sciences (2nd ed.). Monterey, CA: Brooks Cole.

12.

Koenker, R. H. and Hansen, C. W. (1942). Steps for the application of the Johnson-Neyman technique-a sample analysis. The Journal of Experimental Education, 10, 164-173.

13.

Lovell, M. K. , Franzen, M. K. , and Golden, C. J. (1987). Statistical techniques in neuropsychology, IV: Analysis of covariance. The International Journal of Clinical Neuropsychology, IX, 49-55.

14.

Maxwell, S. E. , Delaney, H. D. , and Dill, C. A. (1984). Another look at ANCOVA versus blocking. Psychological Bulletin, 95, 136-147.

15.

Porter, A. C. and Raudenbush, S. W. (1987). Analysis of covariance: Its model and use in psychological research. Journal of Counseling Psychology, 34, 383-392.

16.

Potthoff, R. F. (1964). On the Johnson-Neyman technique and some extensions thereof. Psychometrika, 29, 241-256.

17.

Rao, R. (1952). Advanced statistical methods in biometric research. New York: Hafner.

18.

Wildt, A. R. and Ahtola, 0. (1978). Analysis of covariance. Sage University Paper series on Quantitative Applications in the Social Sciences, series no. 07-012. Beverly Hills and London: Sage Publications.

19.

Winer, B. J. , Brown, D. R. , and Michels, K. M. (1991). Statistical principles in experimental design (3e ed.). New York: McGraw-Hill.