A new class of powerful and informative simultaneous confidence intervals

Area of Science:

• Statistics
• Biostatistics
• Clinical Trial Methodology

Background:

• Step-down tests enhance hypothesis testing power but can yield non-informative simultaneous confidence intervals (SCIs).
• Non-informative rejections in SCIs limit the utility of treatment effect estimates, especially in multi-treatment clinical trials.

Purpose of the Study:

• To develop a new class of SCIs that uniformly improve upon Bonferroni and Holm methods regarding informative rejections.
• To address the deficiency of non-informative rejections in step-down tests when extended to SCIs.

Main Methods:

• Construction of a novel class of SCIs using a dual family of weighted Bonferroni tests.
• Weights in the tests are designed to depend continuously on parameter values.
• Development of a simple algorithm for computation and investigation of shrinkage properties.

Main Results:

• The proposed SCIs uniformly improve upon Bonferroni and Holm SCIs in terms of informative rejections.
• The method demonstrates an attractive shrinkage property for lower confidence bounds.
• The approach is successfully extended to union-intersection tests, including the Dunnett procedure.

Conclusions:

• The new class of SCIs offers a significant improvement for statistical inference in clinical trials, particularly those with multiple treatment arms.
• This method enhances the informativeness of confidence intervals, leading to more reliable treatment effect estimations.
• The findings are validated through simulation studies and illustrated with a real-world clinical trial example.