Recall that latent variables are constructs that cannot be directly measured by the researcher. Public Attitudes toward the police, anxiety, and intelligence are examples. Usually, the researcher will measure multiple indicators as a measure of such constructs. Think about how an IQ test works—lots of questions are used to get at one idea that is summarized as an “IQ Score.” In multiple regression and path analysis, multiple indicators are treated as distinct variables. This causes problems with collinearity. SEMs are superior in this regard. SEM works best with multiple indicators—this is an important part of the measurement submodel. Three or more “good” indicators are preferable.
Last Modified: 02/14/2019