A statistical method for identifying different rules of interaction between individuals in moving animal groups

Area of Science:

• Collective behavior analysis
• Animal social interaction
• Statistical modeling in biology

Background:

• Collective motion emerges from individual rules based on neighbors' positions and behaviors.
• Previous studies aggregated data, overlooking individual differences in interaction rules.
• Qualitative comparisons of individual interaction rules limit understanding of group dynamics.

Purpose of the Study:

• To develop and validate statistical randomization methods for detecting differences in individual interaction rules.
• To quantitatively compare interaction rules between leaders and followers in mosquitofish.
• To establish a reliable framework for analyzing heterogeneity in collective behavior.

Main Methods:

• Development of novel randomization techniques to assess statistical differences in interaction rules.
• Application of these methods to analyze leader-follower dynamics in eastern mosquitofish (Gambusia holbrooki).
• Evaluation of method reliability through repeatability of p-values and consistency in identifying significant differences.

Main Results:

• The proposed randomization methods demonstrate high reliability and repeatability.
• Consistent identification of statistically significant differences in interaction rules between individuals.
• Convergence of test statistics and resolution of ambiguities with increased randomization iterations.

Conclusions:

• The new randomization methods provide a robust, quantitative approach to study individual differences in collective motion.
• These methods are reliable for discerning distinct interaction rules within groups.
• This work advances the quantitative analysis of social interactions and emergent behaviors in animal groups.