Poisson regression for modeling count and frequency outcomes in trauma research

Area of Science:

• Trauma research
• Statistical modeling
• Behavioral science

Background:

• Traditional linear regression may inadequately model count or frequency outcome variables common in trauma studies.
• Count data, such as incidents of aggression or substance abuse, often exhibit non-normal distributions and unequal variances.
• Accurate statistical analysis is crucial for understanding behavioral patterns in trauma research.

Purpose of the Study:

• To demonstrate the application of Poisson regression for analyzing count outcome variables in trauma research.
• To compare the performance of Poisson regression with traditional linear regression for count data.
• To highlight the utility of Poisson regression in studies of intimate partner aggression.

Main Methods:

• Poisson regression analysis was applied to count data representing behavioral incidents.
• Data from a study on intimate partner aggression among male patients in an alcohol treatment program and their partners were used.
• Results from Poisson regression models were compared against those from linear regression models.

Main Results:

• Poisson regression provides a more appropriate framework for modeling count data compared to linear regression.
• The study demonstrated the practical application of Poisson regression in a real-world trauma research scenario.
• Differences in results between the two regression methods were observed, underscoring the importance of choosing the correct statistical model.

Conclusions:

• Poisson regression is a valuable statistical tool for analyzing count outcomes in trauma research.
• The method is particularly useful for understanding the frequency of specific behaviors, such as intimate partner aggression.
• Researchers should consider Poisson regression when dealing with count data to ensure more accurate and reliable findings.