Header image: Architectural widgetsSociology at Duke
Navigation Bar: Resources

Research and Training

Sociological Sites

Computing Resources

Departmental Publications

University Libraries

Quick Links

  
 

LISREL, PRELIS and MULTILEV

Description: These products deal with the methodology of structural equation modeling (SEM). LISREL and PRELIS (a LISREL pre-processor program) were developed by Karl Jöreskog and Dag Sörbom. LISREL is intended for Standard and Multilevel Structural Equation Modeling. The Full Information Maximum Likelihood (FIML) method for missing data is also available for both Standard and Multilevel Structural Equation Modeling. Special methods for the analysis of ordinal variables are available for Standard Structural Equation Modeling of complete data. PRELIS is a application for manipulating, transforming and generating data, computing moment matrices, computing asymptotic covariance matrices, performing regression analyses, performing exploratory factor analyses, performing probit and logistic regression analyses, and performing censored regression analyses. MULTILEV fits multilevel linear and nonlinear models to raw data. MULTILEV is a recent adaption of the LISREL methodology for the analysis of data with hierarchical structures, with new syntax provided to support this type of modeling. The software is developed and marketed by Scientific Software International.

Availability: UNIX versions of LISREL 8.53, PRELIS 2.53 and MULTILEV 5.30 on compute server charisma. SSI offers a free, interactive student Windows edition, that may be downloaded from its website. This product is full-featured with respect to statistical tools, but limited in the size of models it can estimate. The student edition is installed on the PCs in the Room 135 cluster. Full Windows versions of LISREL may be purchased directly from SSI.

Documentation: This product is documented by four manuals, copies of which are kept in Bob Jackson's office.

  • PRELIS 2: User's Reference Guide - describes use of the LISREL preprocessor
  • LISREL 8: User's Reference Guide - describes original syntax for using LISREL
  • LISREL 8: Structural Equation Modeling with the SIMPLIS Command Language - describes preferred, simplified syntax for using LISREL
  • LISREL 8: New Statistical Features - describes new features in LISREL, including multilevel modeling

Additional Help and Information: The LISREL page of the Social Sciences Computing site has pointers to additional web resources.

Modes of Usage Under UNIX: Programs are prepared in advance with an editor (pico, emacs or vi) and stored in your project directory. The convention runs that PRELIS programs are saved with the extension .pr2 and LISREL programs with the extension .ls8. There is also a simplified version of LISREL syntax called SIMPLIS for which programs are typically saved with a .spl extension. The conventional extension for output generated by these applications is .out.

  • To run PRELIS:
    1. On charisma at the UNIX prompt, type: prelis
    2. At the "Input file [INPUT]" prompt, type: commandfile

      [where, commandfile is the name of your PRELIS program]

    3. At the "Output file [OUTPUT]" prompt, type: outputfile

      [where, outputfile is the name of file in which to store results]

  • To run LISREL:
    1. On charisma at the UNIX prompt, type: lisrel
    2. At the "Input file [INPUT]" prompt, type: commandfile

      [where, commandfile is the name of your LISREL program]

    3. At the "Output file [OUTPUT]" prompt, type: outputfile

  • To run MULTILEV:
    1. On charisma at the UNIX prompt, type: multilev
    2. At the "Name of model statements file ?" prompt, type: commandfile

      [where, commandfile is the name of your MULTILEV program]

    3. At the "Name of program output file ?" prompt, type: outputfile

Alternatively, any of these programs may be run in batch mode:

    prelis commandfile outputfile [-size=nnnn] &
    lisrel commandfile outputfile [-size=nnnn] &
    multilev commandfile outputfile [-size=nnnn] &
where, [-size=nnnn] is an optional parameter to specify in Kbytes the maximum amount of dynamically allocatable memory for data storage and estimation. The default allocation is 256 Kbytes.

People Graduate Program Undergraduate Program Resources Home Duke University Home