1
$\begingroup$

From Kutner's Applied Linear Statistical Models

In the next three sections, we shall discuss three multiple comparison procedures for analysis of variance models that permit the family confidence coefficient and the family a risk to be controlled.

Two of these procedures, the Tukey and Scheffe procedures, allow data snooping to be undertaken naturally without affecting the confidence coefficient or significance level.

The other procedure, the Bonferroni procedure, is applicable only when the effects to be investigated are identified in advance of the study.

Why is the Bonferroni procedure "applicable only when the effects to be investigated are identified in advance of the study"?

Is it the same case for the Tukey and Scheffe procedures?

Thanks!

Improve this question
$\endgroup$

1 Answer 1

Reset to default
3
$\begingroup$

Simply because you need to determine the number of tests you're going to correct with the Bonferroni procedure. With the Scheffé & Tukey's HSD procedures this is already determined by the number of possible contrasts & pairwise comparisons respectively.

Improve this answer
$\endgroup$
3
  • $\begingroup$ Then shouldn't they say "in advance of the analysis"? $\endgroup$
    – russellpierce
    Commented Jul 8, 2013 at 23:25
  • $\begingroup$ Everyone decides on the analysis before they conduct the study, don't they ;) $\endgroup$
    – Scortchi ♦
    Commented Jul 9, 2013 at 10:56
  • $\begingroup$ Of course they do. I've never seen anybody do otherwise. They also completely understand why that is crucial to having interpretable p-values. $\endgroup$
    – russellpierce
    Commented Jul 9, 2013 at 22:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.