4. SEM | SPSS AMOS - Introduction to AMOS - Research Coach

IBM SPSS AMOS is built for structural equation modeling (SEM), letting researchers translate a hypothesized theory into a causal path diagram and then test how well the collected data fit that model. AMOS treats relationships as patterns of covariances: it estimates parameters, evaluates goodness of fit, and supports iterative model improvement by modifying or removing paths that don’t align with the data. That workflow matters because SEM is often used to test whether a proposed measurement and causal structure actually matches observed responses.

AMOS’s core modeling distinction is between the measurement model and the structural model. The measurement model links latent constructs (unobservable concepts) to observed indicators (survey items or other measurable variables). The structural model then specifies how constructs relate to one another—typically through hypothesized causal paths. In practice, a researcher draws a diagram that includes both single-headed arrows for directional effects (e.g., an independent/exogenous construct influencing a dependent/endogenous construct) and double-headed arrows for covariances between constructs.

A central concept in AMOS is the latent variable, also called a construct. Latent variables are unobservable by direct inspection—anxiety, motivation, trust, organizational commitment, and job satisfaction are examples—so researchers capture them indirectly using multiple items. Those items (often four to seven, sometimes grouped into dimensions) act as indicators that reflect the underlying construct. In AMOS diagrams, latent constructs are represented as ellipses (ovals/circles), while observed variables/indicators are shown as rectangles or squares.

Because indicators and constructs are measured with imperfection, AMOS incorporates error terms. Each observed indicator carries a measurement error (residual) representing unexplained variance—variance not accounted for by the latent construct. Dependent constructs also include residual terms to represent unexplained variance at the construct level. In AMOS notation, these error/residual terms are treated like unobserved variables and are drawn as circles with one-way arrows.

The transcript also emphasizes terminology overlap that can confuse newcomers: different books and articles may use different labels for the same ideas (e.g., latent constructs, factors, indicators, and measurement items). AMOS’s graphical interface is designed to make these relationships explicit so the conceptual model can be drawn precisely.

Finally, the transcript walks through basic AMOS usage in SPSS AMOS 28 Graphics: selecting observed variables, unobserved variables, and indicators from the toolbar; resizing and deleting shapes; dragging and arranging elements; and using arrows to connect constructs and errors. After building a diagram, running the model produces outputs such as estimates (unstandardized or standardized), loadings, and computation summaries. The workflow culminates in testing hypothesized relationships, with the next step planned as building the first AMOS model.