IBM SPSS AMOS Foundation Course: SEM Scratch to Advanced

Practice a four-variable SEM with internal/external locus of control, well-being as the mediator, and psychological well-being as the outcome, using 50 participants; download data and path diagram from lecture two.

Launch IBM SPSS Amos from the IBM SPSS graphical user interface or as a standalone, and start Amos via the analyze option or from scratch using the Amos plugin.

Explore AMOS graphics tools for building structural equation model and path models, including drawing observed variables, while learning to hover and read tool descriptions to apply them.

Learn that the default model in Amos is any model tested in structural equation modeling, initial measurement model, with other models like covariance and structural weight models listed for analysis.

Explore unstandardized and standardized estimates in SEM with AMOS, learn how B values express unit-driven changes and beta denotes standardized effects in regression paths.

Interpret the model using the computation summary, including cases, model type, default minimization iterations, and chi-square fit, and view listed data sets in the current directory boxes.

Explore the path model canvas to build a structural equation model and use the output tab to view estimates and summaries for interpretation.

Explore the bottom tabs to view a path diagram and tables, showing variables, their estimates, and standard errors via the view text option.

Explore how Amos labels variables in structural equation modeling, distinguishing endogenous, exogenous, observed, latent, and residual variables, and understand tools for drawing and saving observed variables.

A structural equation model abstracts reality, and researchers build a model to approximate that reality using theoretical propositions and hypothesized relationships among variables, including independent and dependent variables.

Recognize exogenous variables as the senders of arrows influencing endogenous variables like wellness, with locus of control internal and external as exogenous predictors, and wellness mediating relationship to psychological well-being.

Understand observed (measured) variables, such as age and gender from questionnaires, and unobserved (latent) variables like personality dimensions, which are inferred through observed items.

Explore residual variables and the error terms E1 and E2 that represent unexplained variance when predicting wellness from locus of control and psychological well-being in regression.

Investigate a structural model of managerial innovation that links search and adapt heuristics and fast and frugal heuristics to heuristic intelligence, which in turn explains business excellence.

Draw an unobserved latent variable and measure it with indicators such as happiness, satisfaction with life, and positive life orientation, in a reflective model with error terms e1, e2, e3.

Use the move object tool to reposition e1, e2, e3 and the latent variable so error terms and observed variables are visible, then apply the touchup tool to remove overflow.

In AMOS, every error term or residual has a built-in constraint: the path value is automatically set to one. This ensures the model remains defined and calculations are possible.

Draw paths with single headed arrows to model relationships in AMOS, treating socioeconomic status as a mediating variable affecting performance, and add error terms for model identification.

Use the draw latent variable tool to create latent and observed variables with residuals in sem, name indicators, and map personality to hard work and life success.

Use the rotate tool in IBM SPSS AMOS to rotate latent-variable indicators by 180 degrees, re-align observed variables, and draw paths from personality to hard work and to success.

Use the erase object tool to delete a selected item from your SEM model in AMOS. Remove its corresponding error term to keep the model consistent.

Moz offers three select tools: select one object, select all objects, and unselect all. Use the move object tool to move the selected objects or the entire model.

Understand how Amos estimates model parameters to minimize the discrepancy between sample and predicted variance-covariance matrices, covering factor loadings, factor variances and covariances, and indicator variances and covariances.

Explore when to apply the maximum likelihood method in sem; the lecture shows that when data are normal, maximum likelihood is the better technique.

Learn when to apply the asymptotic distribution free method, highlighting its use only with very large samples, typically in the thousands.

Explore the maximum likelihood method, which estimates parameters by maximizing the likelihood of the observed data, including factor loadings and error variances, when calculated and predicted parameters align.

Before applying maximum likelihood, verify three assumptions: observed variables are normally distributed and continuous with a large sample; test univariate and multivariate normality.

for non-normal data or many categorical indicators, use partial least squares SEM (PLS-SEM) or asymptotically distribution free (ADF) weighted least squares, which weight observations differentially to improve estimates.

Explore the estimate means and intercepts option to handle missing data by replacing missing values with the mean, a commonly used method when other approaches fail, as in SPSS.

Apply the Israel six correction factor to test your multigroup invariance, ensuring the hypothesized model holds across different groups.

Explore how constraints shape models in Amos, including pre-assigned values, path values, second-order constructs like personality from introversion and extroversion, indicators, and chi-square corrections for many constraints.

Compare fit measures for incomplete data using saturated and independence models against the default model in AMOS to determine the proposed model's adequacy.

Explore how to determine an appropriate sample size for structural equation modeling, with a practical heuristic of 200–300 and guidance to perform sample power analysis for journal or thesis contexts.

Learn how maximum likelihood, generalized least squares, and unweighted least squares in AMOS rely on normality, and discover AMOS's robustness to moderate non-normality while warning against severe deviations.

Learn when regression falls short for complex models and how SEM handles unseen constructs, using personality and hard work to model success and the role of adding variables like IQ.

Explain what exploratory factor analysis is and how it reduces many data items into a few meaningful factors, using personality traits as an example.

Explore how factor analysis, a multivariate technique, uses indicators to define a latent variable, such as personality, and how in structural equation modeling latent and indicator terms recur.

Explore exploratory factor analysis across fields to identify independent factors that explain latent variables, with examples from engineering, psychology, anthropology, and image analysis.

Trace the historical origin of factor analysis from Spearman to Cattell and apply exploratory factor analysis to compress large item pools into meaningful scales and test factors.

Learn how to set up data for exploratory factor analysis in SPSS, import the personality data, and identify meaningful factors that explain variance from 44 ordinal traits.

Perform factor analysis as a data reduction technique, linking latent variables to observed indicators. Use the selection variable to analyze factors across categories such as male and female.

Explore descriptive statistics in SPSS AMOS using univariate descriptives to view the mean and standard deviation for each observed variable, and note the initial solution for commonalities.

Explore building and interpreting a correlation matrix for exploratory factor analysis, including coefficients, significance, and determinant checks, with KMO and Bartlett's tests of sphericity to assess adequacy.

Explore inverse and reproduced correlations, and anti-image analysis to compare correlation and covariance matrices, assess sampling adequacy with KMO and Bartlett's test, and decide on indicator inclusion after factor analysis.

Explore the extraction method in sem amos, focusing on principal component analysis to identify the main component driving data variance, and learn when this method suits research goals.

Learn how maximum likelihood estimation overcomes the sample-population gap in SEM extraction, offering probabilistic estimation and greater generalizability compared to PCA and other methods.

Learn to interpret a correlation matrix and unrotated factor solution in sem, compare rotated and unrotated results, and assess factor loadings, eigenvalues, and significance to determine suitable factor structures.

Compare scree plot and Kaiser's eigenvalue criterion to determine the number of meaningful factors in factor analysis, using unrotated and rotated solutions.

Explore factor rotation in data analysis by clustering many variables into a few meaningful factors using correlations, factor loadings, and orthogonal or oblique rotations.

Compute subject factor scores in SPSS using regression, Bartlett, or Anderson-Rubin methods based on factor orthogonality or oblique assumptions, with default regression and attention to missing values.

Explain the factor score coefficient matrix and its role in calculating factor scores with f = z × b. Show that z scores and beta coefficients determine the factor scores.

Explore missing value analysis in SPSS with three methods: listwise, pairwise, and replace with mean. Compare pairwise and listwise deletion, and learn when to use mean imputation.

Learn to sort the rotated component matrix by size to reveal clean factor groupings and high loadings. Suppress cross loadings by adjusting cutoffs and using Varimax rotation in SPSS AMOS.

Identify latent personality factors by performing a from-scratch factor analysis in SPSS with principal component analysis, scree plot, and varimax rotation.

Conduct reliability analysis with Cronbach's alpha, remove unreliable items by item-total correlations and cross-loadings, and refine factors with principal component analysis until the determinant is non-negative and variance reaches ~61–63%.

Present full factor loadings for unrelated factors, convert the component matrix to APA format, report the sample size, and name factors such as supportive behavior and neuroticism.

Learn to present factor analysis results in APA style by listing factor indicators and loadings, naming six factors (supportiveness, neuroticism, introversion, conscientiousness, artistic, creative), and reporting variance explained with eigenvalues.

Explore how to evaluate factor reliability using Cronbach's alpha, including reverse coding and item deletion, to ensure internal consistency in SPSS AMOS.

Demonstrates improving factor reliability in factor analysis by adding indicators (curious, ingenious, inventive, imaginative, original, reflective) and presenting APA-style tables (Times New Roman 12) with loadings and borders.

Import the EFA model's rotated component matrix into Amos with Pattern Matrix Builder. Specify the data file and perform missing value analysis with EM to validate a six-factor personality scale.

Explore how reliability and validity determine the quality of a model, scale, or psychological construct, and understand how they function as two sides of the same coin.

Assess discriminant validity as part of construct validity by confirming six latent dimensions are distinct and not highly correlated, while evaluating convergent validity to establish overall construct validity.

Explore the concept of validity, showing how a research model or test measures what it claims to measure, using six factors and model-fit and reliability indices to verify adequacy.

Explore convergent validity in SEM with Amos, showing how six sub-dimensions of personality are measured by indicators to ensure the main construct is accurately captured.

Learn how to establish convergent validity using two criteria: average variance extracted above 0.5 and composite reliability above AVE, as outlined by Hare et al. 2013.

Understand how to compute average variance extracted in SEM using standardized factor loadings, and interpret standardized estimates from latent constructs to indicators.

Explore maximum shared squared variance (MSV) and its role in discriminant validity by examining covariances between latent variables and the significance of the absolute value.

Demonstrates that AVE value must exceed MSP value to ensure discriminant validity; when a factor extracts less variance and shares more with another factor, its independence diminishes.

Compute MSV by identifying the maximum covariance a latent variable shares with another factor and squaring it. Use the Validity master plugin to compute MSE and compare with manual results.

Ensure AVE exceeds ESV to establish discriminant validity, since a latent construct should extract more variance than the error term or risk overlap with other constructs.

Explore indices of model fit in structural equation modeling, learning how these metrics judge how well a proposed model matches sample data and potentially generalizes to the population.

compare your proposed model to a baseline null or independent model using incremental fit indices. key indices include cfi, tli, nfi, and ifi.

Report a judicious combination of incremental and absolute fit indices to assess model fit. Emphasize CFI as the incremental index, and include GFI, SRMR, RMK, and chi-square by df ratio.

Describe the chi-square null hypothesis as the comparison of population and model implied covariance matrices, and show good fit when sample, model implied, and population covariances align via maximum likelihood.

Explore the pitfalls of chi-square as a model-fit test in SEM due to size sensitivity, and see why relative chi-square offers a more reliable fit index.

Use relative chi-square value, or chi-square by df ratio, which adjusts for the degree of freedom; aim for 1–3, as 3.6 indicates a poor model fit that needs improvement.

Explore GFI and AGFI as absolute fit measures that reveal how well the model reproduces data; AGFI accounts for degrees of freedom and complexity, with 0–1 range and 0.95 cutoff.

Explore the parsimony based goodness of fit index (PGFI), learn how it penalizes model complexity, and interpret PGFI values relative to GFI when GFI exceeds 0.90.

Calculate srmr with the amos plugin by opening the dialog and clicking calculate estimates. Determine that the default model's srmr is 0.0774, near 0.08 cutoff, so we reject the model.

Explore rmk, a measure of lack of model fit that adjusts for model complexity and provides a 90% confidence interval; near-zero lower limit, upper limit under 0.08, indicate good model.

Compute RMSEA to assess model fit; when RMSEA is 0.075 with a 95% CI above 0.08, the sample covariance and model implied covariance diverge, leading to rejecting the model.