Abstract The significance level and power of analysis of variance and covariance techniques are compared using a computer-simulated experiment. We consider cases assigned to groups by randomization or ‘minimization’, a technique designed to prevent chance imbalance of baseline prognostic variables. The results show that unless analysis of covariance is used for significance testing, the minimization technique can yield invalid tests. Analysis of variance is less powerful with minimization than with randomization when the treatment means are not very different. When analysis of covariance is used, both minimization and randomization will provide an acceptable test of significance, and are more powerful than analysis of variance. In addition, the analysis of covariance, is slightly more powerful with minimization than with randomization. It is suggested that minimization should be considered for group assignment only if all variables used in minimization are also to be used as covariate.
[1]
M. E. Muller,et al.
A Note on the Generation of Random Normal Deviates
,
1958
.
[2]
D R Taves,et al.
Minimization: A new method of assigning patients to treatment and control groups
,
1974,
Clinical pharmacology and therapeutics.
[3]
John P. Gilbert.
Randomization of Human Subjects
,
1974
.
[4]
S. Pocock,et al.
Sequential treatment assignment with balancing for prognostic factors in the controlled clinical trial.
,
1975,
Biometrics.
[5]
M C Weinstein,et al.
Allocation of subjects in medical experiments.
,
1974,
The New England journal of medicine.
[6]
J. Woyciechowska,et al.
Letter: Immunosuppression after vinca alkaloids.
,
1974
.