Ising model
Definition
Ising model is a graphical model that defines a probability distribution over binary variables and is widely used in network psychometrics to model the network structure of psychological variables. The model specifies pairwise interaction parameters that reflect the strength of relationships between variables and threshold parameters that model main effects, with variables represented as nodes and their partial associations represented as edges in the network. Statistical analysis of the Ising model is complicated by an intractable normalizing constant that grows exponentially with the number of variables, making exact maximum likelihood estimation computationally feasible only for small graphs. To overcome this limitation, two approximation methods—joint pseudolikelihood (JPL) and disjoint pseudolikelihood (DPL)—are commonly used in psychometric applications.
Sources: Keetelaar et al. (2024)
Related Terms
Applications
Ising Model and Network Psychometrics
The Ising model is used to represent psychological variables and their conditional dependencies as network structures.
Sources: Keetelaar et al. (2024)
Ising Model and Parameter Estimation
Parameter estimation for the Ising model is challenging due to the intractable normalizing constant; maximum likelihood estimation using exact likelihood is limited for small graphs, while maximum pseudolikelihood estimation methods (JPL and DPL) offer practical alternatives.
Sources: Keetelaar et al. (2024)
