Optimal weighted two-sample t-test with partially paired data in a unified framework

Area of Science:

• Statistics
• Biostatistics

Background:

• Partially paired data presents challenges for standard two-sample t-tests.
• Existing methods for partially paired data have limitations.

Purpose of the Study:

• To develop a unified framework for two-sample t-tests with partially paired data.
• To propose a novel, asymptotically optimal weighted test statistic.

Main Methods:

• Development of a unified framework encompassing existing partially paired t-tests.
• Proposal of a weighted linear combination of test statistics for all paired and unpaired data.
• Performance evaluation through simulation studies.

Main Results:

• The unified framework accommodates various existing two-sample t-tests.
• The proposed asymptotically optimal weighted test statistic demonstrates superior power compared to existing methods.
• The new statistic is applied to real-world datasets in immunology, molecular biology, and sleep science.

Conclusions:

• The unified framework provides a comprehensive approach to analyzing partially paired data.
• The proposed weighted test statistic offers enhanced statistical power for two-sample comparisons.
• The method is applicable to diverse scientific research areas.