Software for Structural Equation Modeling


Here are some suggestions regarding software for doing structural equation modeling.

From: Michael Nielsen[SMTP:mnielsen@gsvms2.cc.gasou.edu]

Sent: Wednesday, January 03, 2001 5:18 PM

To: spsp-discuss@stolaf.edu

Subject: SEM software summary

Thanks to the many people who responded to my query about structural equation modeling software. Six software packages (lisrel, mx, eqs, amos, mplus, and ramona) were recommended for me to consider -- or to avoid, as the case may be! Here are the comments, in a summarized form, grouped mainly by the respondent's preferred software.

- - -

LISREL (For more info, see http://www.ssicentral.com/) The five LISREL advocates emphasized its flexibility. Its detractors most frequently referred to its user interface and the level of sophistication. It is more complicated than other programs, but it can "do anything", including multiple group comparisons and latent growth curves. Advocates also mentioned the PRELIS data screening program, and the SIMPLIS command language for generating easy-to-read output. They also agree that the greater level of sophistication required by the program results in a user who is better able to genuinely understand the analyses being conducted.

- - -

MX (shareware, available at: http://views.vcu.edu/mx/new-features.html) http://views.vcu.edu/mx Two people recommend MX. Their comments were that it is straightforward to program, and seems to be able to do all that LISREL can do. It seems well worth trying.

- - -

EQS (For more info see http://www.mvsoft.com/) EQS and AMOS had an equal number of proponents (N=8) EQS advocates emphasize the program's quick learning curve. "When combined with Barbara Byrne's book, you can hit the ground running (more or less)," was one such comment. (Incidentally, two people strongly endorsed Byrne's book.) EQS doesn't require matrices. Programs can be written like equations or using diagrams. One user, when comparing EQS with LISREL, found some discrepancies between the two. In that instance, LISREL was found to have incorrectly calculated the df in question. Criticisms concern three general areas. One had to do with use support, which was sporadic for some users. The second was with the software being "annoyingly buggy". This seemed most evident when using the model drawing feature. The third criticism was the possibility that it simplifies things too much. (A criticism also leveled at AMOS.) For what it is worth, I tried using the demo version of EQS. It seemed easy to use ... up until the point at which the instructions in the tutorial ask you to do something not shown on the menus. The demo software and manual appear to be slightly different versions, but are different enough that completing the tutorial was more difficult for me than the favorable comments suggested it should be. On the other hand, I like the fact that it comes in windows and mac versions.

- - -

AMOS (See http://www.smallwaters.com/amos/ for more info) AMOS and EQS had an equal number of proponents (8). All advocates of AMOS mentioned its ease of use. Comments such as this were typical: "AMOS is by far the easiest to use. I have found AMOS to be incredibly user-friendly in the last year or so that I have been using it, so I strongly recommend it. I have only an elementary understanding of matrix algebra, but that never prevented me from effectively using AMOS." Additional comments supporting other aspects of AMOS mentioned good customer support, documentation, easy handling of missing data, and ability to export files to SPSS. (Something that appears to be possible with all of the software, as best I can determine, even if it means resorting to a very simple file structure.) The two principle detractors of AMOS suggest that it may be too simple, meaning that the process is simplified to such an extent that it can be possible to run analyses or models that you can't interpret appropriately. A third said that AMOS was too inflexible, with limited options for analyses.

- - -

Mplus

Two responses offered strong support of Mplus. See http://www.statmodel.com They compared it favorably with EQS, but said it had an easier-to-use graphical interface, and that the syntax is superior as well. Model specification is more sensible with Mplus than with other programs. The limited documentation is supplemented by "on-line support [that] is very helpful-actually, it's Linda Muthen [one of the software's authors] who handles all of their queries!" Version 2 is due in February. This is an intriguing piece of software; I'm giving the demo a try.

- - -

RAMONA (SYSTAT 9.0)

Mentioned by one respondent. Said to be very easy to use, but limited in the kinds of analyses it can do.

- - -

Other comments:

One reply recommended an article by Rex Kline, published in Structural Equation Modeling, which reviews/compares the software available at that time.

Software not mentioned, but which may fill people's needs, include the following:

http://cwis.kub.nl/~fsw_1/mto/mto_snw.htm#software

http://www.statsoftinc.com/addition.html#structural

HTTP://hss.cmu.edu/philosophy/TETRAD/tetrad.html

 

Michael Nielsen, Ph.D., Department of Psychology, Georgia Southern University


Back to the Stat Links Page

Visit Karl's Index Page


Contact Information for the Webmaster,
Dr. Karl L. Wuensch



This page most recently revised on 15. May 2001.