Investigating the impact of observation errors on the statistical performance of network-based diffusion analysis

Area of Science:

• Behavioral Ecology
• Social Learning
• Network Analysis

Background:

• Demonstrating social learning in wild populations is challenging.
• Network-based diffusion analysis (NBDA) was developed to infer social learning from observational data using social network structures.
• NBDA models the spread of behaviors, distinguishing between social and asocial learning pathways.

Purpose of the Study:

• To evaluate the performance of NBDA under various conditions.
• To assess NBDA's sensitivity to group size, network structure, observer sampling errors, and diffusion duration.
• To provide guidance for applying NBDA in empirical studies.

Main Methods:

• Simulated trait diffusion through social networks.
• Varied parameters including group size, network heterogeneity, observer sampling error, and time unit aggregation.
• Analyzed Type I error rates in detecting social learning under different conditions.

Main Results:

• Severe observation errors can increase the rate of falsely detecting social learning (Type I errors).
• Aggregating acquisition times into larger time units effectively prevents elevated Type I error rates.
• NBDA demonstrates robustness to sampling errors, performing better than initially anticipated.

Conclusions:

• NBDA is a valuable tool for studying social learning in natural settings.
• The method is resilient to common sources of error in observational data.
• Recommendations are provided for optimizing NBDA application, particularly regarding data aggregation.