Statistical Methods for Unusual Count Data: Examples From Studies of Microchimerism.

Area of Science:

• Human biology
• Immunology
• Biostatistics

Background:

• Microchimerism, the exchange of cells/DNA between mother and fetus, can persist long-term.
• It is linked to various human health outcomes, both positive and negative.
• Analyzing quantitative microchimerism data presents statistical challenges like skewed distributions and excess zeros.

Purpose of the Study:

• To compare statistical models for analyzing quantitative microchimerism data.
• To provide recommendations for best practices in microchimerism data analysis.
• To identify a robust statistical approach for comparing microchimerism levels across groups.

Main Methods:

• Compared Poisson and negative binomial models for quantitative microchimerism data.
• Utilized simulated and observed datasets for model evaluation.
• Modeled microchimerism as a rate (genome equivalents per total cell equivalents).

Main Results:

• Both marginalized zero-inflated Poisson and negative binomial models yielded unbiased estimates.
• The negative binomial model demonstrated accessibility and effectiveness.
• These models facilitate comparisons of microchimerism rates between different groups.

Conclusions:

• The negative binomial model is recommended for analyzing quantitative microchimerism data.
• This approach effectively handles data characteristics and allows for robust group comparisons.
• Proper statistical modeling is crucial for understanding microchimerism's health implications.