Power evaluation of various modified bonferroni procedures by a monte carlo study

Abstract

Various modified Bonferroni procedures (MBPs) have been proposed in order to improve the power of the classical Bonferroni procedure (CBP). In the present paper, powers of these MBPs are investigated by a Monte Carlo study for pairwise comparisons. It is shown that they can be remarkably more powerful than the CBP and Tukey's procedure with respect to all-pairs power, whereas all these procedures with respect to any-pair power are almost the same. Therefore, we recommend the use of MBPs rather than the CBP or Tukey's procedure. Shaffer's procedure sometimes shows higher power than other MBPs and would be the best choice for pairwise comparisons.

A Comparison of Multiplicity Adjustment Strategies for Correlated Binary Endpoints

“…The focus of the evaluation was familywise type I error. In general, the findings parallel those reported for correlated continuous endpoints (Arani and Chen, 1998;Benjamini and Hochberg, 1995;Brown and Russell, 1997;Morikawa et al, 1996;Pocock et al, 1987;Thomas, 1974). Each adjustment procedure examined here protected against the inflation of familywise error seen in unadjusted analyses.…”

“…(1) does not necessarily hold for multiple testing of correlated hypotheses. Several studies have evaluated Bonferroni-type adjustments for comparison of multiple group means (e.g., Benjamini and Hochberg, 1995;Brown and Russell, 1997;James, 1991;Morikawa et al, 1996;Pocock et al, 1987;Thomas, 1974). In the following, we review briefly such multiple comparison adjustments.…”

Independence and Interaction of Regions of the INNER NO OUTER Protein in Growth Control during Ovule Development

“…Significant differences between phenotypic effects of the transgenes were calculated with the modified Bonferroni adjustment (a value/number of pairwise comparisons) to reduce the rate of false-positives while conducting simultaneous pairwise comparisons. This method has been shown to be advantageous when multiple statistical analyses are conducted simultaneously (Wright, 1992;Morikawa et al, 1996).…”

Statistical power of multiplicity adjustment strategies for correlated binary endpoints

“…Relative to the Bonferroni approach, neither alternative sacrifices power. Each of these multiplicity adjustment procedures has been shown to protect against the inflation of familywise error seen in unadjusted analyses [8]. Elsewhere, we have shown that when the correlation, , among endpoints exceeded 0.60, familywise error was unnecessarily conservative with the Bonferroni and Hochberg approaches, typically ranging from 0.02 to 0.04.…”

“…In this context, with alpha thresholds that vary, the Holm and Hochberg procedures can each provide more statistical power than a strict application of the Bonferroni adjustment. The statistical power of strategies for multiple tests of group means has been evaluated [6][7][8][9][10][11]. Yet, none of the aforementioned approaches incorporates the correlations between endpoints.…”