Early stopping for sequential restricted tests of binomial distributions.

Area of Science:

• Biostatistics
• Clinical Trial Design
• Sequential Analysis

Background:

• Restricted sequential procedures can necessitate lengthy trials for null hypothesis acceptance.
• Early indications of null hypothesis unlikelihood may not permit immediate trial termination.

Purpose of the Study:

• To present modifications of restricted sequential plans for binary response trials.
• To enable early stopping with null hypothesis acceptance and reduce expected trial duration.

Main Methods:

• Modifications based on conditional probability of contrary inference.
• Exact properties of stopping-time distributions derived from binomial random walks.
• Analysis of absorption probabilities on a lattice defined by the sequential design.

Main Results:

• Developed tests offer potential for early stopping with null hypothesis acceptance.
• Demonstrated reduction in expected stopping times.
• Calculated relative efficiencies compared to sequential probability ratio tests.

Conclusions:

• Modified restricted sequential procedures enhance efficiency in binary response trials.
• The new methods provide a valuable alternative for clinical trial design.
• Early acceptance of the null hypothesis is feasible, optimizing resource allocation.