Inflation of the type I error rate when a continuous confounding variable is categorized in logistic regression...

Area of Science:

• Biostatistics
• Epidemiology
• Statistical Modeling

Background:

• Adjusting for confounding variables is crucial in statistical analysis to prevent biased results.
• Categorizing continuous variables can simplify analysis but may introduce analytical issues.
• Type I error inflation is a concern in hypothesis testing, leading to false positives.

Purpose of the Study:

• To investigate the inflation of the type I error rate in statistical significance testing.
• To assess the impact of adjusting for a categorized continuous confounding variable on a continuous risk factor.
• To identify factors influencing this type I error inflation.

Main Methods:

• Utilized Monte Carlo simulation methods for empirical assessment.
• Simulated scenarios involving a continuous risk factor and a correlated continuous confounding variable.
• Varied sample size, correlation strength, and number of categories for the confounder.

Main Results:

• Demonstrated a significant inflation of the type I error rate under the specified conditions.
• Observed that error inflation increases with larger sample sizes.
• Found that higher correlation between the risk factor and confounder exacerbates the issue.
• Error inflation was more pronounced with fewer categories for the confounding variable.

Conclusions:

• Testing the significance of a continuous risk factor after adjusting for a categorized confounder can lead to inflated type I error rates.
• The extent of type I error inflation is dependent on sample size, correlation, and the degree of categorization.
• Researchers should be cautious when using categorized confounders, especially in large studies with strong correlations.