Regression analysis of biomedical research data based on a repeated measure or cluster sample

Area of Science:

• Biostatistics
• Medical Research Methodology

Background:

• Biological and medical studies frequently use cluster samples where data are repeatedly measured from the same subject or related subjects.
• Standard regression analysis often ignores cluster identity, leading to biased statistical inference and results that misrepresent the average within-cluster effect.

Purpose of the Study:

• To present a robust multiple regression model for analyzing cluster sample data.
• To provide a method that yields valid statistical inference for the average within-cluster regression coefficient.

Main Methods:

• A multiple regression model is proposed that incorporates cluster identity.
• Clusters are treated as a nominal confounding variable, represented by a set of dummy variables within the regression model.

Main Results:

• The proposed dummy variable multiple regression model provides an accurate estimate of the average within-cluster regression coefficient.
• This approach ensures valid statistical inference, overcoming the limitations of ignoring cluster identity.

Conclusions:

• The dummy variable multiple regression model effectively analyzes cluster sample data in biology and medicine.
• This statistically sound method is accessible through standard statistical software packages.