Variable Selection in the Regularized Simultaneous Component Analysis Method for Multi-Source Data Integration.

Area of Science:

• Data Science
• Statistics
• Interdisciplinary Research

Background:

• Interdisciplinary research frequently integrates data from multiple sources for comprehensive analysis.
• Techniques like regularized simultaneous component analysis (regularized SCA) are used for joint analysis of diverse datasets, such as GPS and travel diary data.
• Effective variable selection is critical for regularized SCA to identify joint and unique sources of variation.

Purpose of the Study:

• To compare the performance of six different variable selection methods for regularized SCA.
• To identify the most effective variable selection method for analyzing integrated datasets in interdisciplinary studies.

Main Methods:

• The study simulated data with varying noise and sparseness levels.
• Six variable selection methods were evaluated: cross-validation (CV) with the "one-standard-error" rule, repeated double CV (rdCV), BIC, Bolasso with CV, stability selection, and Index of Sparseness (IS).
• Performance was assessed based on the ability to select appropriate variables for regularized SCA.

Main Results:

• The Index of Sparseness (IS) demonstrated superior performance compared to the other five methods.
• IS proved to be a computationally efficient and effective variable selection technique.
• The findings highlight the importance of method selection for successful data integration.

Conclusions:

• The Index of Sparseness (IS) is recommended as the preferred variable selection method for regularized SCA.
• This research contributes to improving data integration techniques for complex interdisciplinary projects.
• Accurate variable selection enhances the reliability and interpretability of results from joint data analysis.