Path analysis is used commonly to evaluate direct and indirect associations among observed variables. Although there are sufficient user guides for commonly referenced structural equation modeling (SEM) software programs, there is little information to help the researcher compare usability and outcomes of these programs. Therefore, deciding which SEM software program to use often presents a challenge for both novice and experienced researchers.
To evaluate path analysis results from one data set using two commonly referenced SEM programs, AMOS (Analysis of Moment Structures) Version 6.0 and LISREL (Linear Structural Relations) Version 8.80.
Using one data set, each researcher worked solely within one SEM program and was blinded to the solution obtained by the other until analyses were complete. Subsequently, each researcher replicated the other's solution in the opposite program. Finally, a reduced model using modification indices was obtained by each researcher and then compared across SEM programs for similarity.
Results showed extremely similar but not totally identical solutions. Modification indices suggested one additional variable and two extra paths in both AMOS and LISREL.
An examination of the solutions obtained in the two SEM programs suggested small discrepancies in the critical ratios due to rounding, which may have produced subtle alterations in the order of path deletion during model reduction. Despite minor differences, the final solutions could be replicated accurately in each program. Researchers should select an SEM program based on their programming knowledge and the research questions being addressed. Regardless of the program selected, the researcher can have confidence in the comparability of results.
Margaret F. Clayton, PhD, RN, FNP-CS, is Assistant Professor; and Marjorie A. Pett, MStat, DSW, is Research Professor, College of Nursing, University of Utah, Salt Lake City.
Accepted for publication December 10, 2007.
Corresponding author: Margaret F. Clayton, PhD, RN, FNP-CS, College of Nursing, University of Utah, 10 South 2000 East, Salt Lake City, UT 84124 (e-mail: firstname.lastname@example.org).