Rank regression: an alternative regression approach for data with outliers

Area of Science:

• Health Services Research
• Biostatistics
• Mental Health Research

Background:

• Linear regression models are prevalent in mental health and health services research.
• Classic linear regression assumes normal data distribution, which is often unmet in practice.
• Non-normal data and outliers can compromise the reliability of standard regression analyses.

Purpose of the Study:

• To introduce and evaluate rank regression as an objective method for handling non-normal data with outliers in health research.
• To compare the performance of rank regression against classical and semi-parametric regression models.

Main Methods:

• Utilized simulated and real-world datasets containing non-normal distributions and outliers.
• Applied classical linear regression, semi-parametric models, and rank regression techniques.
• Compared the reliability and objectivity of estimates generated by each regression method.

Main Results:

• Classical and semi-parametric models produced unreliable estimates when data included outliers.
• Rank regression demonstrated a more objective approach to handling outliers and non-normal data.
• Rank regression yielded more robust and trustworthy results compared to traditional methods.

Conclusions:

• Rank regression is a valuable and objective alternative for analyzing health services data with non-normal distributions and outliers.
• This method enhances the reliability of statistical findings in mental health and related research where data often deviates from normality.