Simulated Intervention on Cross-Sectional Nested Data: Development of a Multilevel NIRA Approach

2025-06-27

AI Summary

Existing node identification algorithms (NIRAs) cannot accommodate cross-sectional nested data, limiting their applicability to causal inference in multilevel psychological networks. To address this, we propose multilevel NIRA (ml-NIRA), the first extension of NIRA to nested structures with within-group correlation. By integrating psychological network modeling with a multilevel statistical framework, ml-NIRA enables cross-level intervention simulation. It explicitly models hierarchical dependencies while preserving network sparsity and estimating inter-level causal effects. We systematically formalize ml-NIRA’s modeling workflow, parameter estimation strategy, and interpretive framework for intervention effects, validating its performance via simulation studies and empirical analysis. This work substantially broadens the scope of causal inference in psychological networks, offering a reproducible and interpretable computational tool for mechanistic investigation in multilevel contexts—such as education and clinical settings—while clarifying key limitations, including sensitivity to level imbalance and model misspecification.

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Social Networks and Social Media: Computational social science
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• User Modeling, Personalization and Recommendation: Psychology-informed user models and recommender systems

Abstract

With the rise of the network perspective, researchers have made numerous important discoveries over the past decade by constructing psychological networks. Unfortunately, most of these networks are based on cross-sectional data, which can only reveal associations between variables but not their directional or causal relationships. Recently, the development of the nodeIdentifyR algorithm (NIRA) technique has provided a promising method for simulating causal processes based on cross-sectional network structures. However, this algorithm is not capable of handling cross-sectional nested data, which greatly limits its applicability. In response to this limitation, the present study proposes a multilevel extension of the NIRA algorithm, referred to as multilevel NIRA. We provide a detailed explanation of the algorithm's core principles and modeling procedures. Finally, we discuss the potential applications and practical implications of this approach, as well as its limitations and directions for future research.

Problem

• Identifies causal relationships in cross-sectional nested data
• Extends NIRA algorithm to multilevel network analysis
• Addresses limitations in simulating directional psychological networks

Innovation

• Extends NIRA to multilevel nested data
• Simulates causal processes cross-sectionally
• Details algorithm principles and modeling