Empirically derived dietary patterns: interpretability and construct validity according to different factor rotation...

Area of Science:

• Nutrition Science
• Statistical Modeling

Background:

• Dietary patterns are crucial for understanding diet-health relationships.
• Factor analysis is commonly used to derive dietary patterns.
• The choice of factor rotation method can influence pattern results.

Purpose of the Study:

• To evaluate how different factor rotation methods affect the interpretability and construct validity of dietary patterns.
• To compare orthogonal (varimax) and oblique (promax, direct oblimin) rotations.
• To assess the impact of factor loading cut-offs (≥ |0.20| and ≥ |0.25|) on derived dietary patterns.

Main Methods:

• Exploratory factor analysis was used to derive dietary patterns in 1,102 Brazilian adults.
• Varimax, promax, and direct oblimin rotations were applied.
• Confirmatory factor analysis and goodness-of-fit indexes were used to assess construct validity and model fit.

Main Results:

• Differences in dietary pattern composition and interpretability were noted between varimax and promax/oblimin at a ≥ |0.20| cut-off.
• These differences diminished at a ≥ |0.25| cut-off.
• No patterns derived at ≥ |0.20| showed acceptable model fit, while promax rotation at ≥ |0.25| demonstrated the best fit.

Conclusions:

• Factor rotation methods and factor loading cut-offs interact to influence dietary pattern outcomes.
• A factor loading cut-off of ≥ |0.25| with promax rotation is recommended for robust dietary pattern analysis in this population.
• The findings highlight the importance of methodological choices in dietary pattern research.