Analyzing k (2 x 2) tables under cluster sampling

Area of Science:

• Biostatistics
• Epidemiology
• Longitudinal Data Analysis

Background:

• Multiple measurements within subjects (e.g., periodontal studies) create data dependence.
• Standard statistical methods assuming independence are inappropriate for correlated outcomes.
• The Mantel-Haenszel statistic's distribution is affected by this correlation.

Purpose of the Study:

• To propose a modified Mantel-Haenszel procedure for analyzing dependent data.
• To provide a statistical method that adjusts for correlation between observations within clusters.
• To develop a non-iterative technique accommodating site-specific data.

Main Methods:

• A modified Mantel-Haenszel statistic is developed based on generalized estimating equations.
• The method assumes no specific correlation structure within clusters.
• It offers a closed-form, tabular adjustment technique.

Main Results:

• The proposed method provides a valid adjustment for correlated observations.
• It allows for the inclusion of site-specific exposure and covariate information.
• Demonstrates applicability using a periodontal study example.

Conclusions:

• The modified Mantel-Haenszel procedure effectively handles correlated data in clustered studies.
• This approach enhances the accuracy of association analysis when observations are not independent.
• Applicable to various fields with longitudinal or clustered data, such as ophthalmology and periodontal research.