A unified approach to multiple-set canonical correlation analysis and principal components analysis

Area of Science:

• Statistics
• Data Analysis
• Psychology

Background:

• Canonical correlation analysis (CCA) and principal components analysis (PCA) are common data reduction techniques.
• CCA describes associations among variable sets, while PCA maximizes variance in a single set.
• Existing methods have distinct objectives, limiting their combined application.

Purpose of the Study:

• To develop a unified framework integrating CCA and PCA for data reduction.
• To provide a flexible approach allowing a compromise between maximizing inter-set associations and explaining variance.
• To offer a novel method for analyzing complex datasets in fields like psychology and neuroimaging.

Main Methods:

• A unified framework is proposed, treating CCA and PCA as special cases.
• A single optimization function is developed, combining criteria from both CCA and PCA.
• Analytical minimization of the optimization function is performed for parameter estimation.

Main Results:

• The unified approach successfully integrates CCA and PCA.
• The method allows for solutions that balance explaining variance within datasets and maximizing associations between datasets.
• Simulation studies and functional neuroimaging data analysis demonstrate the approach's utility.

Conclusions:

• The proposed unified framework offers a powerful and flexible tool for data reduction.
• This approach enhances the analysis of multivariate data by combining the strengths of CCA and PCA.
• The method has broad applicability, particularly in psychological research and neuroimaging analysis.