A latent variable represents a construct that is unobserved and not directly measurable. Latent variables are often hypothetical or abstract concepts. In the behavioral sciences, examples of such variables might include quality of life, depression, and selective attention. In contrast, measures such as a single neuropsychological subtest score or a behavioral rating scale score are observed variables. Such a measure can also be referred to as an indicator variable, as it indicates or reflects the latent construct. Whereas a single indicator variable measures one dimension of behavior, the latent variable can collectively explain a wider range of behaviors or processes.
Some statistical analyses, such as structural equation modeling (SEM), use latent variables as a representation of the shared variance (or covariance) of multiple indicator variables. Although observed measures inevitably entail measurement error, latent variables account for this error and provide a clearer...