Efficient Exploration of Many Variables and Interactions Using Regularized Regression.

Area of Science:

• Prevention Science
• Machine Learning
• Statistical Modeling

Background:

• Prevention science studies frequently encounter challenges like high dimensionality, multicollinearity, and covariate selection issues.
• Conventional statistical methods often prove inadequate for analyzing complex datasets in prevention research, potentially compromising study validity and statistical power.
• Regularized regression techniques, while established, are underutilized in prevention science.

Purpose of the Study:

• To introduce regularized regression as a powerful machine learning tool for prevention science.
• To demonstrate the utility of regularized regression in analyzing complex datasets common in prevention research.
• To highlight how regularized regression can overcome limitations of traditional methods.

Main Methods:

• Application of regularized regression, a machine learning technique, to a real-world dataset.
• Utilized data from the Youth Risk-Behavior Surveillance System (YRBSS) 2015.
• Analyzed a dataset comprising 76 variables (151 including interactions) with a sample size of 7979.

Main Results:

• Regularized regression effectively handled a dataset with numerous variables and interactions.
• The technique facilitated the exploration of complex data structures, yielding insights.
• Demonstrated the feasibility and benefits of applying regularized regression in prevention science contexts.

Conclusions:

• Regularized regression is a valuable tool for prevention researchers dealing with high-dimensional and complex data.
• Its application can improve the analysis of prevention science data, enhancing statistical power and validity.
• Encourages wider adoption of regularized regression in the field for deeper data insights.