The statistical consequences of repeated significance testing of follow-up dental data were evaluated by computerized Monte Carlo simulation. Substantial distortions in nominal alpha level were noted, particularly when the scores for individual subjects were not highly correlated between statistical observations. Suggested methods of controlling this distortion include the pre-specification of a terminal statistical analysis that would be given best evidence status and the partitioning of alpha error among all analyses by Bonferroni's inequality.
Get full access to this article
View all access options for this article.
References
1.
Armitage, P. (1975): Sequential Medical Trials, 2nd ed., Oxford, U.K.: Blackwell Scientific Publications, pp. 145-146.
2.
Goodson, J.M.; Tanner, A.C.R.; Haffajee, G.C.; and Socransky, S.S. (1982): Patterns of Progression and Regression of Advanced Destructive Periodontal Disease, J Clin Periodontol9:472-481.
3.
Haffajee, A.D.; Socransky, S.S.; and Lindhe, J. (1983): Comparison of Statistical Methods of Analysis of Data from Clinical Periodontal Trials, J Clin Periudontol10:247-256.
4.
Nemenyi, P. (1978): Linear Hypotheses. III. Multiple Comparisons. In: International Encyclopedia of Statistics, Vol. 1, Kruskall, W.H. and Tanur, M., Eds., New York, New York : The Free Press, p. 569.
5.
Tukey, J.W. (1977): Some Thoughts on Clinical Trials, Especially Problems of Multiplicity, Science198:679-684.
