The Importance and Statistical Procedures of Priori Comparisons among Means Author: Der-shin Fan

Abstract：

The analysis of variance (ANOVA) currently enjoys the status of being probably the most uSed (some would say abused) statistical technique in every field of research. In the past, when this procedure was used, from the very begining researchers usually ran an overall F-test. If a significant F value was obtained, what they would have shown was simply that the overall null hypothesis was false and continue to run post hoc comparisons to investigate hypothesis involving means of individual groups or sets of groups. On the other hand, If the overall F value was not significant, then the procedure was brought to an end. While post hoc comparisons are important in actual research, especially for exploring one's data and for getting leads for future research, the method of priori comparisons is perhaps more powerful and of greater scientific value.
The purposes of this article are : (1) to describe the meanning and importance of priori comparisons and related error rate and (2) to introduce the priori comparison procedures of multiple t test, orthognoal contrasts, and Dunn's test.

《Full Text》