ABSTRACT
The process of specifying a structural equation model can be complex, regardless of the soundness of a conceptual model, data quality or sample size. Not all models lead to meaningful numerical solutions. Identifiability is a condition under which it is possible to determine a unique solution (i.e. unique set of parameter estimates). This chapter presents the necessary conditions for model identification as well as illustrative examples of underidentified, just identified and overidentified models. Overidentified models, with positive degrees of freedom, provide some meaning to model fit measures. This chapter also discusses equivalent models, which are different models with identical model fit. Empirical methods cannot help to distinguish which equivalent model fits the data best. Ultimately, a team science approach can be useful for modifying models that are not identified and choosing between equivalent models in studies using structural equation modeling.
