Network Psychometrics

Network psychometrics models psychological constructs as networks in which observed items are nodes and edges are partial correlations between items after controlling for all other items.

네트워크 심리측정은 관측 문항을 노드로, 다른 모든 문항을 통제한 문항 간 편상관을 연결선으로 하는 네트워크로 심리적 구성개념을 모형화하는 방법이다.

How do I analyze how individual symptoms or questionnaire items are connected to each other?

개별 증상이나 설문 문항들이 서로 어떻게 연결되어 있는지 어떻게 분석하나요?

Network psychometrics estimates a regularized partial correlation network among items, typically a Gaussian graphical model, so that each edge shows the unique association between two items. It fits symptom or attitude data where the interest is in item-level connections and central items rather than a single latent factor.

When to use Network Psychometrics

• Mapping connections among symptoms, attitudes or questionnaire items
• Identifying central or bridge items within or across scales
• Comparing network structures between groups

When to use something else

• You assume a common latent cause behind the items: use confirmatory factor analysis
• You need the dimensional structure of the items for scoring: use exploratory factor analysis

Assumptions and requirements

• Adequate sample size relative to the number of nodes
• Items without heavy redundancy (near-duplicate content)
• Appropriate correlation type for the data (e.g., polychoric for ordinal items)
• Cross-sectional networks describe between-person associations, not causal dynamics

What the analysis reports

• Network graph with weighted edges
• Centrality indices (strength, expected influence)
• Edge-weight and centrality stability (bootstrap, correlation-stability coefficient)
• Community detection results

How to report it

The estimated network contained [n] non-zero edges out of [total] possible, and [item] showed the highest strength centrality (CS-coefficient = [value]).

Related methods

• Confirmatory Factor Analysis (CFA)
• Exploratory Graph Analysis
• Gaussian Graphical Model

Run Network Psychometrics on Quorant