Three p-value consistent procedures for multiple comparisons with a control in direction-mixed families.

Area of Science:

• Statistics
• Clinical Trials
• Biostatistics

Background:

• The Dunnett procedure is standard for comparing multiple treatments to a control.
• Existing methods struggle with mixed one- and two-sided hypotheses and p-value consistency.

Purpose of the Study:

• To develop p-value consistent single-step and stepwise statistical procedures.
• To offer more powerful and reliable methods for clinical trial analysis.

Main Methods:

• Development of novel single-step and two stepwise procedures.
• P-value consistency and average power were evaluated via simulation.
• Comparison with existing methods like the Hochberg step-up procedure.

Main Results:

• Proposed stepwise procedures are more powerful than single-step methods.
• New procedures demonstrate robustness to distribution changes.
• The proposed step-up procedure outperforms the Hochberg approach.

Conclusions:

• The new procedures provide consistent and powerful statistical testing for clinical trials.
• Tabulated critical values facilitate practical implementation.
• These methods enhance the analysis of multiple treatment comparisons.