Lisrel 8.8

Strukturgleichungsmodelle bieten als multivariate Forschungsmethode die Möglichkeit komplexer Datenanalysen. Sie verbinden konfirmatorische Faktorenanalysen mit der linearen Regression und erlauben so die Analyse latenter Strukturen. Hypothetische Konstrukte werden in diesem Ansatz als latente Variablen aufgefaßt. Sie werden mittels mehrerer Indikatoren operationalisiert. So ist es möglich, die Meßfehler der einzelnen Indikatoren zu bestimmen und die „fehlerfreien“ regressiven Beziehungen zwischen den hypothetischen Konstrukten zu analysieren, was irreführend auch als Kausalanalyse bezeichnet wird. Strukturgleichungsmodelle werden in lineare Gleichungssysteme umgesetzt. Die unbekannten Parameter können unter bestimmten Voraussetzungen aus den beobachteten Daten geschätzt werden. Eine besondere Rolle spielt dabei die Maximum-Likelihood-Methode, die gleichzeitig die globale Überprüfung eines Strukturgleichungsmodells anhand eines c2-Wertes ermöglicht, der sich aus dem Minimum der Fitfunktion ergibt. (aus Lisrel Einführung von Lars Satow). Today, however, LISREL for Windows is no longer limited to SEM. The latest LISREL for Windows includes the following statistical applications. LISREL for structural equation modeling. PRELIS for data manipulations and basic statistical analyses. MULTILEV for hierarchical linear and non-linear modeling. SURVEYGLIM for generalized linear modeling. CATFIRM for formative inference-based recursive modeling for categorical response variables. CONFIRM for formative inference-based recursive modeling for continuous response variables. MAPGLIM for generalized linear modeling for multilevel data. The PRELIS, LISREL and SIMPLIS manuals (as PDF) are included with the LISREL program.

New features in LISREL 8.8 for Windows

• Structured latent curve models
  The LISREL CO command has been extended to include the exponential (EXP) and natural logarithm (LOG) operators as well as parentheses. This allows LISREL users to fit, for example, the structured latent curve models outlined in Browne (1993).
  
  
• Factor analysis of ordinal variables
  Classical exploratory factor analysis assumes that the observed variables are continuous. The PRELIS OFA command implements exploratory factor analysis of ordinal variables as described in Jöreskog & Moustaki (2006).
  
  
• Generalized linear models (GLIMs) for multilevel data
  The new statistical application MAPGLIM fits generalized linear models to multilevel data. Users can select from the multinomial, Bernoulli, Poisson, binomial, negative binomial, Normal, Gamma and inverse Gaussian sampling distributions. The corresponding link functions include the log, cumulative logit, cumulative probit, complementary log-log and logit link functions.
  
  
• Observational residuals
  Bollen and Arminger (1991) introduced observational residuals for structural equation models. LISREL 8.8 for Windows allows users to compute observational residuals along with latent variable scores for the latent variables of the model. This implementation is described and illustrated in Jöreskog, Sörbom & Wallentin (2006)
  
  
• Writing parameter estimates, standard error estimates and measures of fit to a PSF
  The PV, SV and GF keywords on the LISREL OU command or the SIMPLIS LISREL output command have been extended to allow users to save the parameter estimates, standard error estimates and measures of fit to a PSF. This is especially useful for Monte Carlo studies.
  
  
• Changes to the graphical user interface (GUI)
  The main window of LISREL 8.8 for Windows is now entitled LISREL for Windows. The revised Export Data option on the File menu of the main window allows users to export data to various data formats such as SPSS, SAS, SYSTAT, Statistica, etc.
  
  

New features in LISREL 8.7 for Windows

• Generalized Linear Models (GLIMs) for complex survey data
  The new SurveyGLIM module in LISREL 8.7 allows users to select from the multinomial, Bernoulli, binomial, Poisson, negative binomial, gamma, Gauss, and inverse Gaussian sampling distributions. Various link functions, such as the log, cumulative logit, cumulative probit, complementary log-log, and logit are available.
  SurveyGLIM allows for the analysis of data from a simple random sample or from a complex sample design. In the latter case it is assumed that the population from which the sample is obtained can be stratified into strata. Within each stratum, clusters (primary sample units or PSUs) are drawn and within each stratum-cluster combination, the ultimate sampling units (USUs) are drawn with specified design weights. There is also an option to correct for finite populations, provided that the sampling rates or population sizes are available.
  
  
• Implementation of design weights in the LISREL Multilevel modeling module
  There has been a growing interest in recent years in fitting models to data collected from surveys using complex sample designs. LISREL 8.7 features an option for users to include sample design weights for the analysis of hierarchical linear models. This makes it possible to specify weights on levels 1, 2 or 3 of the hierarchy. Correct parameter estimates and robust standard errors are produced under complex sampling designs.
  
  
• Implementation of sampling weights for SEM models when data is missing at random
  In previous versions of LISREL, users were able to compute the appropriate covariance and estimated asymptotic covariance matrices for continuous variables via PRELIS given a normalized weight variable. These matrices are only produced in the case of complete data, or using list-wise deletion in situations where missing data values are present.
  In version 8.7, it is possible to use design weights to fit SEM models to continuous data with missing values. The easiest way to do this is to define the weight variable once a PSF file is displayed. A full information maximum likelihood (FIML) method is used to obtain the correct parameter estimates and robust standard errors given the sampling weights.
  
  
• Multivariate Censored Regression
  Univariate regression for a censored response variable is available since LISREL 8.54. In LISREL 8.7, this method is extended to allow for multivariate censored regression. In addition, the appropriate sample covariance matrix for a set of censored variables may be computed and used to fit structural equation models to censored data.
  
  
• Goodness-of-fit statistics
  Since the release of LISREL 8.52 for Windows, the computation of the chi-square test statistic value for the independence model is based on the normal-theory weighted least squares (NT-WLS) chi-square test statistic value rather than on the minimum fit function chi-square test statistic value. This change implied that the goodness-of-fit statistics, which is based on the chi-square test statistic value for the independence model such as the CFI, NFI, NNFI, IFI, etc., were different and led to numerous inquiries by our LISREL users. As a result, LISREL 8.7 produces an additional file with the file extension “FTB” that contains a listing of these goodness-of-fit statistics based on all four chi-square test statistic values that LISREL 8.7 reports.
  
  
• Changes to the windows/menus/dialogs
  There are three new options in the Compute dialog box starting with version 8.7 of LISREL. These are: (i) TIME (ii) AUTOLAG/ORDER, (iii) CHISQ(DF)
  The first option enables users to create a new variable called TIME, that assumes integer values 1, 2, 3, …, ncases. Functions of TIME, for example TIME**2 can also be computed. The second option allows the user to create new variables that assumes the same values than an existing variable, but with a user-specified lag. These new variables are useful in identifying time series processes and for the calculation of lagged correlation matrices. Lastly, one can generate random deviates from a chi-square distribution with a specified number of degrees of freedom.
  Additions/changes to the dialog boxes of the multilevel module include: (i) No-Intercept option (ii) Select weights list box (iii) Print asymptotic covariances checkbox (iv) Print values of within and between covariance matrices checkbox. Note that the specification of a level-1 ID variable is no longer required.
  
  

New statistical features introduced in 8.50

• Structural Equation Modeling with incomplete data: Efficient Full Information Maximum Likelihood (FIML) for incomplete data that are missing at random.
• Multilevel Structural Equation Modeling with complete and incomplete data.
• Nonlinear Multilevel Modeling: Two-level nonlinear regression models.
• Exploratory Data Analysis: Formal Inference-based Recursive Modeling (FIRM) for detecting complex statistical relationships among categorical and continuous variables.
• Multiple Imputation: Expected Maximization (EM) or Markov Chain Monte Carlo (MCMC) for imputing incomplete data that are missing at random under the assumption of an underlying multivariate normal distribution.
• PRELIS System Files: LISREL 8.50 for Windows uses a PRELIS System File (*.psf) to store information such as number of observations, number of variables, variable names, type of variable, category labels, missing value codes and the raw data. When opened the .PSF file is displayed in spreadsheet format and a PSF toolbar appears which enables users to make graphic displays, define and compute variables and analyze the data. A *.psf file can now also be specified as part of the LISREL or SIMPLIS syntax. Use of a *.psf file greatly facilitates the ability to draw path diagrams and build syntax interactively.
• External Data Sources: Import data from numerous external data sources with no limitations on the number of observations and variables, except for those imposed by the computer resources.
• Additional Graphical Displays: Pie charts, Box-and-Whisker plots and matrix scatterplots.
• Windows Interface: A 32 bit Windows interface for Windows 95, 98, 2000 and ME supporting long path and file names. Note that Windows 3.1 and 3.11 are no longer supported.
• Documentation: Revised online Help file covering all new features and a 500 page user's guide describing the LISREL user interface, new statistical features and syntax

System Voraussetzungen:

Weitere Informationen:

• Lisrel homepage
• Lisrel online help