Weighted mean difference statistics for paired data in the presence of missing values

Area of Science:

• Biostatistics
• Biomedical Data Analysis
• Statistical Methodology

Background:

• Missing data is prevalent in biomedical studies, particularly with paired designs.
• Partially paired data, arising from various issues like loss to follow-up, complicates statistical comparisons.
• Existing methods struggle with the complexities introduced by missing values in paired data.

Purpose of the Study:

• To propose a novel class of statistical tests for comparing distributions with partially paired data.
• To develop an optimal weight for this test statistic to maximize its performance.
• To evaluate the efficacy of the proposed test against existing methods using simulations and real-world data.

Main Methods:

• A general class of test statistics based on the difference in weighted sample means was developed.
• No specific distributional or model assumptions were imposed on the data.
• An optimal weight was derived for the proposed test statistic.

Main Results:

• Simulation studies demonstrated that the proposed test with the optimal weight performs well.
• The new method outperformed existing statistical approaches in practical scenarios.
• The test was successfully illustrated using two cancer biomarker studies.

Conclusions:

• The proposed statistical test offers a robust solution for analyzing partially paired biomedical data.
• The optimal weight enhances the test's performance, making it a valuable tool for researchers.
• This method provides a reliable approach for comparing variables between two conditions when data is incomplete.