Psychometrics using SPSS and AMOS

Learn how to download and install IBM SPSS, register for a 14‑day trial, choose 64‑bit or 32‑bit, and locate the program files on your system.

Download and install AMOS 24 trial on Windows from the IBM site, sign up with an IBM ID if needed, and complete the setup in the program files x86 folder.

Explore reliability analysis as a fundamental statistical tool in psychometrics, and learn to report the reliability and validity of scales used in your study for valid conclusions.

Assess reliability as the consistency of a test or scale in measuring a true score across time and items, highlighting the role of internal consistency and validity.

Explore how a scale uses a global construct T reflected by indicators X1 to Xk, with factor loadings and error terms in a reflective measurement model, unlike formative scales.

Explain how Cronbach's alpha is best for unidimensional constructs and why composite reliability matters for first-order and second-order (hierarchical) models, with caution on multidimensional scales.

Assess test-retest reliability by correlating scores from two administrations of the same test. A 15-day to one-month gap minimizes learning effects and yields meaningful stability estimates.

Parallel form reliability creates two test forms, an original and an equivalent version, and uses their correlation to estimate reliability.

Analyze internal consistency reliability, its use of Cronbach's alpha, and why it links indicators to a single construct beyond test-retest and parallel forms.

Discover how Cronbach's alpha measures internal consistency by showing how test indicators relate and converge on a global construct across Likert-type, dichotomous, and other scales.

Examine two core assumptions of Cronbach's alpha: uncorrelated error terms and tau-equivalent indicators. Note that correlated errors imply redundancy, and distinguish tau-equivalent from congeneric models for reliable scale interpretation.

Explore the Cronbach's alpha formula using variance and correlation methods, with practical insight from SPSS, detailing items, indicator variances, and the total composite variance.

Explore Cronbach's alpha interpretation for reliability, including the acceptable 0.7+ range, 0–1 bounds, and cautions about values above 0.9 signaling possible multicollinearity; aim for 0.7–0.9.

Assess Cronbach's alpha on the personality scale using the 44-item dataset in SPSS, and explore 'scale if item deleted' to identify items to delete to improve reliability.

Assess the case processing summary: 433 valid cases (94.3%) from 459 subjects, with 26 exclusions (5.7%). Note Cronbach's alpha is 0.54 across 44 items, below the acceptable 0.7.

Diagnose and handle missing data in SPSS by recoding system missing values to -999, then run reliability analysis to evaluate Cronbach's alpha and improve scale reliability.

Diagnose scale reliability by analyzing the scale mean and covariances to identify problematic items. Explore item statistics, standard deviations, and mid-range means to improve Cronbach's alpha for a five-point scale.

Diagnose scale reliability using the item total table to identify items with low or negative item-total correlations and delete those to improve Cronbach's alpha.

Eliminate items with negative item-total correlation, then those with low item-total correlation (below 0.4); re-run reliability analysis to raise Cronbach's alpha to an acceptable range.

The item discrimination index shows how well an item differentiates high- and low-scoring individuals via item total correlations. Items with a total correlation above 0.4 indicate good discrimination.

Explore factor analysis, a data reduction technique that compresses many traits into a few core factors, illustrated by reducing thousands personality adjectives to a handful.

Explore factor analysis as a multivariate technique that uses indicators of a latent variable to form manageable factors, and learn the terms latent variable and indicator.

Explore how factor analysis, including exploratory factor analysis, clusters observations into independent factors that explain latent variables across social science and engineering, with real research examples.

Trace the historical origin of factor analysis from Spearman to Cattell and Allport, and show how exploratory factor analysis compresses thousands of adjectives into test factors.

Explore the differences between exploratory factor analysis and confirmatory factor analysis. Learn to use SPSS for exploratory and confirmatory factor analysis, and other modeling tools like Amos, Mplus, or Lisrel.

Set up the data for exploratory factor analysis in SPSS, using a 44-attribute personality dataset to identify meaningful factors that explain variance on an ordinal scale.

Learn how factor analysis, a data reduction technique, reduces many indicators into latent dimensions, distinguishing latent variables from observed indicators, and how to use a selection variable for subgroup analyses.

Explore univariate descriptives in SPSS, including mean and standard deviation, and review the initial solution details to identify potential variable issues for psychometrics analysis.

Analyze correlation matrices for exploratory factor analysis, learning how coefficients, significance, and determinant shape factorability, and how KMO and Bartlett's tests assess sampling adequacy and matrix suitability.

Explore how the inverse, reproduced, and anti-image matrices are used in SPSS factor analysis, compare correlation and covariance, and interpret KMO and Bartlett tests for sampling adequacy.

Explore the extraction method in factor analysis, focusing on principal component analysis, its use for identifying the main component in data, and its historical critique and applications.

Explore principal axis factoring to identify independent, uncorrelated factors in factor analysis, compare with principal component analysis, and understand how variance is distributed across factors.

Use the maximum likelihood extraction method to avoid PCA and principal axis factoring limitations that treat the sample as the population, yielding probability-based, generalizable results.

Choose between correlation and covariance matrices for factor analysis based on measurement scales and commensurability. Prefer correlation for interpretable factors, and transform to z-scores when using a covariance matrix.

Explore interpreting correlation matrices and unrotated factor solutions, examining factor matrices, eigenvalues, factor loadings, rotation, significance, and criteria like 0.3 loading thresholds to assess multicollinearity and unidimensionality.

Compare scree plot with the eigenvalue criterion to decide how many factors to extract, using both since eigenvalues over one indicate meaningful factors and the scree plot can be unclear.

Rotation in factor analysis clusters many variables into a few meaningful factors by rearranging axes according to correlations, revealing either orthogonal or oblique factor structures via loadings.

Explore rotation methods in SPSS, including varimax, quartimax, equamax, direct oblimin, and promax, and compare orthogonal versus oblique rotations for interpretable factor loadings.

Compute subject factor scores in SPSS using regression, Bartlett, or Anderson-Rubin methods to match orthogonal or oblique factors, using factor loadings, z scores, and subject scores.

Examine the factor score coefficient matrix and its use in calculating overall factor scores, with f equal to z times b, where z is a z score.

Use missing value analysis in SPSS: pairwise, listwise, and mean replacement. Pairwise uses available data, listwise drops any missing case, and mean preserves all observations.

Sort by size clarifies how rotated component loadings cluster variables into factors, by choosing a cutoff (0.3–0.6) to minimize cross-loadings and refine factor structure in SPSS factor analysis.

Perform factor analysis on a personality dataset from scratch in SPSS, identify independent personality factors with principal component analysis, and interpret KMO, Bartlett's test, communalities, eigenvalues, and rotated loadings.

Correct negative determinants by performing reliability analysis with Cronbach's alpha and removing unreliable items. Iteratively refine the factor solution using principal component analysis and rotation to maximize explained variance.

Identify dimensions of personality using factor analysis with SPSS and AMOS, show full loadings and APA-formatted component matrices, and report n=459. Name factors as supportive behavior and neuroticism.

Learn to present factor analysis results in APA style, listing factors with their loading indicators. Six factors named from indicators include supportiveness, neuroticism, introversion, lack of conscientiousness, artistic, and creative.

Explore how to compute Cronbach's alpha for factor reliability in SPSS, using reliability analysis, reverse coding, and 'scale if item deleted' to assess internal consistency across four factors.

Learn how to present factor analysis results in APA style, evaluate reliability, and optimize factor indicators using SPSS and AMOS.

Import the exploratory factor analysis model into AMOS using the rotated component matrix and pattern matrix builder to validate a six-factor personality scale, missing value analysis and reliability checks.

Assess reliability and validity as two sides of model quality, applicable to scales, research models, or any psychological construct. Understand how they together define quality in constructs and models.

Understand validity as the ability of a test or model to measure what it claims, with AMOS indices evaluating reliability and validity of the six-factor personality model.

Explore construct validity and convergent validity in structural equation modelling with AMOS, validating a personality construct through six subdimensions, latent variables, and their indicators.

Assess convergent and discriminant validity within a six-factor personality model using AMOS to confirm distinct latent constructs and overall construct validity.

Average variance extracted (AVE) measures the variance captured by indicators from the latent construct relative to error terms in a reflective measurement model. An AVE above 0.5 indicates convergent validity.

Learn how to calculate average variance extracted (AVE) from standardized factor loadings in structural equation modeling. See how squaring standardized weights and dividing by the number of indicators yields AVE.

Learn to manually calculate AVE in Excel by squaring standardized weights, averaging across indicators, and compare with the AMOS master validity tool by James Gaskin to assess model fit.

Learn about maximum shared square variance (MSV) and its role in discriminant validity by measuring the maximum squared covariance a latent variable shares with another, using absolute values.

Explain why AVE must exceed MSV for discriminant validity, defining AVE as variance a latent factor explains in its indicators and illustrating how shared variance with other factors reduces independence.

Learn to manually compute the maximum shared variance (MSV) for each factor by squaring its largest covariance with another latent variable, and verify results against the validity master plug-in.

Explore average shared square variance (ASV) as a measure for discriminant validity, calculating the average of squared variances shared between a latent construct and others, and contrast with MSV.

Analyze why AVE should exceed ASV to establish discriminant validity, showing that a latent construct captures more variance than it shares with other constructs.

Master manual ASV calculation by squaring a construct’s covariances with other factors, averaging them, and comparing ASV to AVE to assess discriminant and convergent validity in a six-factor model.

Explore indices of model fit in structural equation modeling, showing how fit measures compare model behavior to data, how sample size affects them, and their implications for population generalization.

Distinguish incremental and absolute fit indices and apply typical cutoffs using AMOS, with examples such as CFI, TLI, NFI, NNFI, IFI, GFI, AGFI, SRMR, and RMSEA.

Incremental fit indices measure how much a proposed model improves over a baseline, null, or independent model, using CFI, TLI, NFI, NNFI, and IFI.

absolute fit indices measure a structural equation model's ability to reproduce the given dataset, with closer-to-reality models yielding better fit and more distant models yielding poorer fit.

Report a judicious mix of incremental and absolute fit indices, prioritizing CFI, and include GFI, SRMR, RMSEA, and chi-square by df ratio for model evaluation.

Learn to locate and interpret model fit indices in AMOS by running the analysis, viewing the AMOS output, and comparing default, saturated, and independent models to determine model quality.

Explore CMIN as a chi-square goodness-of-fit test in AMOS, comparing sample and population covariance matrices to assess model fit and the null hypothesis.

Express the null hypothesis of the chi-square goodness-of-fit test as Σ = Σ(Θ) and compare sample, model-implied, and population covariance matrices to assess model fit and discrepancy.

Explore why chi-square tests can mislead SEM model fit due to large sample sensitivity, and learn to use relative chi-square to compare model fit with degrees of freedom.

Assess model fit using the relative chi-square (chi square by df ratio) to adjust for sample size. Interpret values between 1 and 3 as acceptable; 3.6 indicates poor fit.

Explore the goodness of fit index (GFI) and adjusted GFI (AGFI), absolute fit measures that show how well a model reproduces sample data and penalize model complexity.

Explore how PGFI, a parsimony based goodness of fit index, penalizes model complexity; with GFI above 0.90, PGFI is typically below 0.50 and AGFI below 0.50.

Explore SRMR, the standardized root mean square residual, and RMSEA as discrepancy measures in AMOS, using residuals between sample and model-implied covariance matrices to assess model fit.

Compute SRMR in AMOS using the standardized SRMR dialog, interpret a 0.0774 SRMR against the 0.08 cutoff, and understand why a poorly fitting model may be rejected.

Explore RMSEA as a key measure of lack of model fit in structural equation modelling, showing its adjustment for model complexity and its 90% confidence interval alongside SRMR.

Calculate RMSEA from the model fit view; an RMSEA of 0.075 indicates poor fit between the sample covariance S and the model implied covariance, leading to rejection of the model.

Explore plug-ins in AMOS, including built-in and user-installed tools, and learn how to use them for tasks like testing tool validity and identifying model fit measures.

Locate AMOS plug-ins by navigating to program files (x86) on a 32-bit system, then SPSS, AMOS, and plug-ins to view draw covariances, master validity, and model fit; copy plug-ins here.

Learn how to locate and access the AMOS 24 plug-ins folder by navigating to AppData in the user profile, enabling hidden items, and opening the AMOS 24 plug-ins directory.

Discover how to access and download AMOS plug-ins from Statswiki, unzip them, and organize a dedicated 'AMOS plug-ins' folder for efficient model validation in AMOS.

Install four Professor Gaskin plugins in AMOS 23 or lower, including common latent factor, master validity, model fit measures, and pattern matrix builder, by copying dll files and unblocking them.