Visualizing results from infection transmission models: a case against "confidence intervals"

Area of Science:

• Epidemiology
• Mathematical Modeling
• Data Visualization

Background:

• Stochastic transmission models are crucial for understanding infectious disease dynamics.
• Communicating the large datasets generated by these models presents significant challenges.
• Current visualization methods, like mean/median with 95% intervals, can lead to misinterpretation.

Purpose of the Study:

• To propose novel visualization techniques for stochastic model outputs.
• To address the limitations of traditional methods in representing complex data.
• To enhance the communication of epidemic model results, including rare events.

Main Methods:

• Development of two alternative visualization approaches for stochastic model results.
• Focus on conveying central tendency (median) and uncertainty (95% interval).
• Incorporation of methods to highlight outlier or unexpected model outcomes.

Main Results:

• The proposed methods effectively display median and 95% intervals.
• Alternative visualizations provide clearer insights into stochastic model behavior.
• These approaches better capture the range of potential epidemic scenarios, including extreme events.

Conclusions:

• Novel visualization techniques offer improved clarity for stochastic epidemiological models.
• These methods mitigate ambiguity and potential misinterpretation associated with standard plots.
• Enhanced visualization aids in understanding and communicating the full spectrum of epidemic model predictions.