[A comprehensive introduction to LISREL models, including path analysis, structural equations, and confirmatory factor analysis, with many worked examples using the LISREL and EQS programs. This is a moderately difficult, graduate level text, but it is the most complete treatment of LISREL models available. The discussion of rules and methods for identifying parameters of structural equation and factor analysis models is excellent.]
[ This book was designed specifically as a starting point for people who always wanted to use LISREL, but were too afraid to try. Byrne provides input and output for 6 different types of applications - 3 using single samples and 3 using multiple samples. The data are taken from her research, so each is accompanied by the referenced article for those who may want to follow up on the substantive aspects of the topic. The examples are somewhat marred by the use of poor quality printouts for the input and output.]
[As the title indicates, a relatively elementary text, at the graduate or advanced undergraduate level. The last few chapters contain a good discussion of designing a factor analytic study, with a case study of Comrey's personality scales.]
[A nicely integrated collection of papers by various authors illustrating the application of CFA and SEM models in psychology and behavioural science. An appendix includes the LISREL statements for the analyses in all the chapters.]
[If brevity is the soul of wit, Everitt gets the prize for this compact, readable treatment of exploratory and confirmatory factor analysis in under 70 pages, of which half is devoted to a variety of LISREL models. Numerical examples are provided for most of the methods discussed and an additional chapter describes latent variable models for categorical data.]
[A simplified introduction to the history and basic ideas of factor analysis from Spearman through Thurstone.]
[A graduate-level text, but not too difficult mathematically. Contains many worked examples, and excellent coverage of the practical issues in conducting factor analytic research.]
[A substantial revision of Gorsuch (1973), with extensive comments on using computer programs and a new chapter on confirmatory factor analysis.]
[A simple non-technical introduction to the ideas of factor analysis for social scientists. The "Coles Notes" of factor analysis.]
[Long considered "the bible" of factor analysis methods, Harmon covers all the classic methods of extracting and rotating factors, with a great many worked examples. No coverage of CFA or developments post 1966.]
Focuses on using CALIS for CFA and SEMs under SAS. Each chapter describes a plausible research context, input and output for PROC CALIS, and interpretation of results. Chapter 6 covers SEM.
An introduction focusing on AMOS.
[A non-technical introductory text for LISREL. The major focus of the book is on causal submodels. Discusses estimation problems.]
[A practical, detailed treatment of principal components analysis and related methods. In addition to the usual topics one would expect, the book covers multidimensional scaling and preference analysis, correspondence analysis, applications of PCA to regression and MANOVA, robust PCA, with a brief treatment of factor analysis.]
[An excellent introductory-level presentation of modern methods for analysis of covariance structures and structural equation modelling. A particularly readable, non-technical (less than Everitt or Long) introduction to recent developments in the field including path diagrams, confirmatory factor models, simplex, test theory, multitrait-multimethod models and some others.]
[A non-technical introduction to the aims of confirmatory factor analysis.]
[Discusses test/measurement theory ideas, e.g., reliability in the context of factor models and item response theory (i.e., latent trait models).]
[A comprehensive graduate-level text which covers the mathematical basis for methods of factor analysis from the early Thurstonian methods up through Joreskog's development of confirmatory methods.]
[Chapter 6 of this book is a fascinating account of the role played by the early history of factor analysis in Cyril Burt's attempt to establish a hereditary, causal explanation of intelligence in terms of Spearman's g factor. In Gould's description he gives a clear non- technical account of the aims of factor analysis and the dangers of giving factors a wealth of theoretical meaning only because they came out of some mathematical process. Gould's treatment is not particularly balanced, but his arguments are generally sound.]
[Compared Horn's method, Scree test, Bartlett's chi², Kaiser's eigenvalue > 1 and Velicer's minimum average partial correlation method. Kaiser's method tended to severely overestimate the number of components. Horn's and Velicer's methods generally performed quite well.]
[Illustrates the use of LISREL for testing various forms of factorial invariance across groups. They demonstrate procedures for identifying noninvariant measurement parameters and testing differences in latent factor means. When groups are found to differ in some parameters (e.g., factor loadings), they demonstrate the use of LISREL methods to pinpoint the source of the differences.]
[Describes difficulties (negative variance estimates, failure to converge) encountered with estimation of some models for MTMM data -- those which postulate equal loadings of variables on the trait and method factors -- which they ascribe to the model being unidentified. Several alternative ways to model MTMM data by CFA are suggested.]
[How can you establish whether a given test measures the same trait dimension, in exactly the same way, in two or more distinct groups of individuals? This paper compares the utility of CFA and item response models used to investigate whether mood ratings collected in Minnesota and China were comparable.]
[Surveys over 70 applied studies using structural equation models.]
[Considered a "classic" nontechnical introductory text on the topic.]
[AMOS is an IBM/PC program for Analysis of Moment Structures, including all the models handled by LISREL, EQS and SAS PROC CALIS. The latest DOS/Windows version has a menu-driven front end which makes it quite easy to use; a Windows version provides a graphical interface (AMOS Draw) which allows you to specify a model by drawing the path diagram rather than using matrices or linear equations. When you estimate the model, the coefficients are shown on the path diagram.The AMOS program has the best facilities for testing and comparing multiple models for the same set of data, and for multi-sample analyses: much easier to set up, and the output provides all the model-comparison statistics described by Bollen (1989) and Marsh, Balla, & McDonald (1988). The program is also unique in providing facilities for bootstrapped estimates of standard errors. The User's Guide contains an extensive set of worked examples, which provide an excellent tutorial introduction to the use and interpretation of structural equation and CFA models. Further information is contained on the Amos Home Page]
[Discusses the use of the chi² test in ACOVS and LISREL, as well as goodness of fit indices used to compare models, including the Tucker-Lewis index.]
[A CFA example, using data from Wheaton (1978) on psychological disorders of patients over two time periods. Shows the setup and results from the LISREL and EQS programs.]
[Describes some major problems in applying and interpreting covariance structure analysis.]
[The CALIS Procedure (Covariance Analysis and LInear Structural equations) is the SAS answer to LISREL. CALIS is far more flexible than LISREL (it provides 5 different ways to specify models), but is, in some ways, somewhat more complex. This extended user's guide is now largely incorporated in the SAS/STAT manual, but contains some additional technical details and an extensive list of sample applications, which are all available in the SAS/STAT Sample Library.]
[Describes the formulation of assessing inter- rater reliability in terms of test-theory constructs of parallel tests and shows how these may be estimated and tested as confirmatory factor models. Standard methods based on the intraclass correlation coefficient assume compound symmetry, while these models do not. Two worked examples are included.]
[Joreskog's first paper applying the ACOVS model to test-theory questions and multitrait-multimethod matrices. Most of the examples are reprinted in Joreskog (1974).]
[A somewhat technical collection of papers that introduced the LISREL framework. Discusses all LISREL submodels including factor mean comparisons across populations.]
[This manual for the LISREL program contains a wide variety of illustrative examples, with background, input and output, for the various special cases of the LISREL model, including structural equations (path) models, confirmatory factor analysis, multi-sample models, and models with means structures. See SPSS (1990) for details specific to SPSS.]
[Presents a comprehensive tutorial on the use of LISREL to test models of first- and second-order factors and factorial invariance across groups. The data analyzed comes from 28 subscales of the Self-Description Questionnaire used to examine components of self-concept in 658 children in grades 2 - 5. LISREL specifications are given for a number of models tested.]
[Examines more than 30 indices which have been proposed for testing the goodness of fit of LISREL & ACOVS modesls, using real and simulated data. The Tuker-Lewis (1973) index was the only widely-used index that was relatively uninfluenced by sample size. Contrary to claims by Joreskog & Sorbom (1981), their GFI and AGFI indices were influenced by sample size.]
[A sensitive discussion of the trade-off between goodness-of-fit and parsimony in structural equation models. Describes how to adjust the LISREL GFI index by a measure of parsimony for a model.]
[With version 7, LISREL has become more like a regular SPSS procedure, and the syntax more SPSS-like. PRELIS is a pre-processor for LISREL. Among other things, it computes polyserial and polychoric correlations necessary for a proper analysis of categorical or discrete data, as in item analysis.]
[Examines 12 factor models for a battery of conventional tests computerized adaptive versions designed to measure the same aptitudes, using a double cross- validation design. A good example of analysis of multi- trait, multi-method data. The factor model provides an estimate of the (disattentuated) method correlation between conventional and adaptive testing.]
[Rotter believed his Internal-External scale was unidimensional, i.e., that a single general factor could explain most of the correlations among the items. Marsh & Richards summarize 20 published factor analysis studies of the scale which show that 4-6 factors are required. They carry out a confirmatory analysis of a five-factor model, with the factors allowed to be correlated. A second-order model tested whether the correlations among the first-order factors could be accounted for in terms of a single second order factor; this model provided a good fit to the data, and the second-order factor was interpreted as a generalized I-E construct. The paper contains a good tutorial discussion of the use of goodness-of-fit indices in comparing different models for covariance structure.]
[A nice example of the use of CFA for studying convergent and divergent validity via analysis of the MTMM matrix.]
© 1995 Michael Friendly
Author: Michael Friendly