21. SPSS AMOS Series | First/Basic Structural Model in AMOS
Based on Research With Fawad's video on YouTube. If you like this content, support the original creators by watching, liking and subscribing to their content.
Load the dataset in AMOS before drawing latent variables and indicators for the constructs under study.
Briefing
A basic AMOS structural model can be built to test whether authentic leadership significantly predicts life satisfaction—using one latent independent variable, one latent dependent variable, and a clear set of indicators plus error terms. The workflow starts by loading the dataset, then drawing two latent variables: “Authentic Leadership” with five indicators and “Life Satisfaction” with its corresponding indicators. Indicators are arranged so the independent variable sits on the left and the dependent variable on the right, and the layout is adjusted while preserving AMOS “symmetries” so the diagram remains structurally consistent.
Once the measurement portion is in place, the model is completed by adding error terms for the unobserved variables. AMOS initially creates generic error terms (E1 through E10), which are then renamed to match the constructs—specifically labeling the errors associated with authentic leadership and life satisfaction. A single directional path is then drawn from authentic leadership to life satisfaction, forming the core structural relationship being tested. The diagram is resized and the canvas orientation can be switched to landscape if needed to fit the model cleanly on the screen.
Running the model requires selecting appropriate analysis properties, including standardized estimates, modification indices, correlations, and R-square outputs. After the first run, AMOS may flag an issue: endogenous variables can appear without residual error variables. In this case, an additional unique variable (an error term) is added to the endogenous variable “Life Satisfaction” so the model aligns with the expectation that endogenous constructs should have residual error associated with them.
After saving the model (named “FM first model”), the output is checked in two layers. First, model fit statistics are reviewed—CFI, TLI, GFI, and the chi-square divided by degrees of freedom (with RMS EA noted as not very good). The fit is described as “pretty good,” with the RMS EA weakness attributed to earlier simplifications for the sake of demonstrating the structural-model process (including how modification indices and error covariances were handled).
Second, the structural effect is evaluated. The path from authentic leadership to life satisfaction shows a standardized estimate around 0.501 and an unstandardized regression weight around 47. The critical ratio exceeds 1.96 and the p-value is significant, indicating the relationship is statistically supported. With these significance checks, the conclusion follows: authentic leadership has a significant impact on life satisfaction. The session ends by framing this as the standard process for building, running, and interpreting a first structural model in AMOS—especially when measurement models and structural models share the same indicator structure and (in later work) the same error covariances.
Cornell Notes
The session walks through building a first/basic structural equation model in AMOS to test whether authentic leadership predicts life satisfaction. It starts by loading the dataset, drawing two latent variables (authentic leadership with five indicators and life satisfaction with its indicators), and arranging them left-to-right. Error terms are added and renamed so each construct has appropriate residuals, then a single path is drawn from authentic leadership to life satisfaction. After selecting analysis options (including standardized estimates and R-square), the model is run and fit indices are reviewed. Finally, the structural path is judged significant using critical ratio (>1.96) and p-values, leading to the conclusion that authentic leadership significantly influences life satisfaction.
What is the minimum structural setup for testing one independent and one dependent variable in AMOS?
Why does AMOS require residual error terms for endogenous constructs?
How are model fit statistics used in this structural-model run?
What specific output values determine whether the structural path is significant?
What practical steps help keep the AMOS diagram readable without breaking the model?
Review Questions
- What changes are made after the first model run when AMOS reports endogenous variables with no residual error variables?
- Which two statistics are used to decide whether authentic leadership significantly predicts life satisfaction, and what thresholds are referenced?
- How do CFI, TLI, GFI, and chi-square/df contribute to judging the model before interpreting the structural path?
Key Points
- 1
Load the dataset in AMOS before drawing latent variables and indicators for the constructs under study.
- 2
Create two latent variables: one for the independent construct (authentic leadership) and one for the dependent construct (life satisfaction), then attach the correct number of indicators to each.
- 3
Arrange indicators so the independent latent variable sits on the left and the dependent latent variable sits on the right, and preserve symmetries when moving objects.
- 4
Add and rename error terms (unobserved variables) so constructs have appropriate residuals, especially for the endogenous variable.
- 5
Draw a single path from authentic leadership to life satisfaction to represent the structural hypothesis being tested.
- 6
Run the model with standardized estimates and R-square outputs, then check fit indices (CFI, TLI, GFI, chi-square/df, RMS EA) before interpreting the path.
- 7
Use p-values and critical ratios (with the referenced 1.96 benchmark) to conclude whether the structural relationship is statistically significant.