Step-up multiple testing of parameters with unequally correlated estimates

Area of Science:

• Biostatistics
• Statistical Inference
• Multiple Hypothesis Testing

Background:

• Simultaneous hypothesis testing requires controlling the familywise error rate (FWER) to prevent false discoveries.
• Existing step-up procedures often assume equal correlations among parameter estimates, limiting their applicability.
• Unequal correlations arise in practical scenarios, such as experiments with varying sample sizes across groups.

Purpose of the Study:

• To extend the Dunnett and Tamhane (1992a) step-up multiple test procedure to accommodate unequally correlated parameter estimates.
• To enhance the robustness and applicability of multiple testing procedures in biopharmaceutical research and other fields.
• To compare the performance of step-up versus step-down multiple testing approaches.

Main Methods:

• Development of an extended step-up procedure for simultaneous hypothesis testing with unequal correlations.
• Theoretical analysis to ensure familywise error rate control under the new assumptions.
• Comparative analysis of step-up and step-down methods.

Main Results:

• The proposed extension successfully incorporates unequally correlated parameter estimates into the step-up testing framework.
• The method maintains the specified familywise error rate alpha.
• Demonstration of applicability to biopharmaceutical testing problems with unequal sample sizes.

Conclusions:

• The extended step-up procedure offers a more flexible and powerful approach for simultaneous hypothesis testing when correlations are unequal.
• This advancement is particularly relevant for complex experimental designs common in drug development and clinical trials.
• The study provides valuable insights into choosing between step-up and step-down testing strategies.