Understanding Diagram Symbols in SEM Model

SEM uses diagrams to denote relationships to be tested. It is important that you understand what these diagram symbols mean because AMOS is going to make you draw out your concep- tual model. One of the frustrating aspects of SEM is that there are often multiple terms that mean the exact same thing.You can read three different books, and all three can use different terms to mean the same thing. Hence, I have consolidated all the overlapping terms here to add greater clarity to the different nomenclature.

1. Latent Variable/Construct—A latent variable is also referred to as an “unobservable”. This is a concept that cannot be directly It would be nice to just look at someone and tell his/her level of anxiety, but the fact is some people are great at hiding their true feelings. In these instances, we cannot simply observe a person and determine levels of anxiety. Thus, concepts such as anxiety are an “unobservable” and require the researcher to find a way to capture the concept by other means, such as asking survey questions.You will also see the term “Factors” used when referring to latent/unobservable constructs. Examples of unobservable constructs in psychology are Anxiety, Motivation, and Trust. In sociology, examples are Perceived Stress and Disconnectedness. Other areas, such as business, have constructs like Brand Attitude, Intentions to Purchase, Employee Burn- out, and Satisfaction. Examples of education constructs are Frustration, Self-Efficacy, and Engagement. Examining unobserved constructs has even expanded into diverse areas such as wildlife and fisheries, with constructs like Attitudes toward wildlife/pollinators and Intentions for conservation practices. These unobservable constructs are often measured by “indicators”, or “measurement items”, which often take the form of survey questions. For example, you might ask survey questions (indicators) to measure a consumer’s level of trust with a company (unobservable).

The diagram shape represented for latent/unobservable variables is a circle or oval.

2. Observed Variable/Indicator—As the name denotes, measures are taken to capture an unobservable concept through observable This can be done through survey questions, manipulations, or behavioral tracking.This concept is also referred to as Mani- fest Variables or Reference Variables.You will often see the term “items” and “indicators” used to represent observed variables. In essence, observed variables/indicators are the raw data captured that will be used to explain concepts in a SEM model. These observed variables/indicators can be categorical, ordinal, or continuous.

The diagram shape represented for observed variables is a square or rectangle.

3. Measurement Error/Residual Term—Measurement error represents the unex- plained variance by an indicator measuring its respective latent construct. In trying to capture an unobservable construct with a measurement indicator, the unexplained vari- ance in the measurement is error, or “measurement error”. Along with the indicator, an error term is also present on a dependent latent construct. This is the unexplained variance on the construct level as a result of the relationships from the independent Error terms for latent variables are also called residual terms or disturbance terms. Since measurement error and residual terms represent unexplained variance, AMOS treats these terms like unobserved variables. Thus, the symbol for these error terms are a circle and a one-way arrow.

The diagram shape represented for error terms is a circle and one-way arrow.

4. Direct Path Effect—Hypothesized directional effects of one variable on

The diagram shape is represented by a line and a single arrowhead.

5. Covariances—the amount of change in one variable that is consistently related to the change in another variable (the degree to which the two variables change together on a reliable and consistent basis).

The diagram shape for a covariance is a curved line with two arrowheads.

Source: Thakkar, J.J. (2020). “Procedural Steps in Structural Equation Modelling”. In: Structural Equation Modelling. Studies in Systems, Decision and Control, vol 285. Springer, Singapore.