Understanding data in clinical research: a simple graphical display for plotting data (up to four independent...

Area of Science:

• Clinical Research
• Biostatistics
• Data Visualization

Background:

• Logistic regression analysis (LRA) is increasingly used in clinical research for dichotomous outcomes.
• Current LRA methods lack easily interpretable graphical displays, often relying on complex logit functions.
• Effective data visualization is crucial for understanding statistical analysis in clinical settings.

Purpose of the Study:

• To propose novel, understandable graphical display techniques for binary logistic regression analysis.
• To enhance the interpretation of LRA results for clinical researchers and physicians.
• To provide methods for visualizing both simple and complex LRA models.

Main Methods:

• Transforming the logit function into a logistic function to derive predicted probabilities P(Y=1|X).
• Generating 2D plots for binary LRA with a single predictor variable.
• Proposing a 3D surface graphic for visualizing models with up to four independent variables (two continuous, two discrete).

Main Results:

• The proposed methods allow for the transformation of complex logit functions into easily graspable figures.
• A simple 2D plot can represent LRA with one predictor.
• A 3D surface graphic is presented as a method to visualize LRA with multiple predictors.

Conclusions:

• The developed graphical techniques offer a more intuitive understanding of binary LRA in clinical research.
• These visualizations can improve data interpretation and knowledge acquisition for medical professionals.
• The proposed methods are accessible and do not require sophisticated statistical packages, utilizing standard 2D or 3D plotters.