Testing similarity in longitudinal networks: The Individual Network Invariance Test.

Area of Science:

• Psychological network analysis
• Structural equation modeling
• Individual differences research

Background:

• Comparing idiographic network structures to assess heterogeneity is difficult.
• Previous methods like visual inspection or multilevel modeling do not directly test network equality.
• Existing techniques lack the ability to directly compare individual network structures.

Purpose of the Study:

• Introduce the Individual Network Invariance Test (INIT) for comparing idiographic networks.
• Implement INIT within an R package for practical application.
• Provide a direct method for testing the equality of idiographic network structures.

Main Methods:

• Developed the Individual Network Invariance Test (INIT).
• Implemented INIT in an R package.
• Evaluated INIT using a simulation study with varying network structures (saturated and pruned) and sample sizes.
• Utilized chi-squared difference tests and model selection criteria (AIC, BIC).

Main Results:

• INIT demonstrates adequate performance with 100 time points per individual.
• For saturated networks, the Akaike Information Criterion (AIC) was the best model selection criterion.
• For pruned networks, the Bayesian Information Criterion (BIC) showed superior performance.
• An empirical example demonstrated INIT's utility for testing network equality within and across time.

Conclusions:

• INIT provides a statistically valid method for testing the equality of idiographic network structures.
• The choice of model selection criterion (AIC vs. BIC) depends on whether networks are saturated or pruned.
• INIT enables direct comparison of individual networks, advancing the study of individual differences.