7
$\begingroup$

I have a question regarding the two-step method for performing a structural equation modeling problem.

The two-step method indicates that you should

  1. First construct the Measurement Model to ensure its accuracy
  2. Once 1. is set, you can estimate the structural paths by simultaneously evaluating the measurement and structural portions.

My question is this:

Do I get a Measurement Model set in stone, then proceed to estimate the entire model all together? And when reporting this, do I report the measurement parts that I obtained from step 1, or from step 2?

Thanks.

$\endgroup$
2
  • 2
    $\begingroup$ Anderson and Gerbing (88) proposed the two-step method, and then muliak (00) introduced the four-step method for evaluating SEM. I have yet to see any postings regarding this matter. $\endgroup$
    – CoryB
    Commented Jul 18, 2015 at 3:36
  • $\begingroup$ In most papers I have read the fit of the measurement model is reported without structural parameters, followed by an interpretation of the structural model parameters. $\endgroup$
    – s_suzuki
    Commented Dec 26, 2020 at 4:53

1 Answer 1

1
$\begingroup$

Do I get a Measurement Model set in stone, then proceed to estimate the entire model all together?

Yes, their idea was to first fix measurement-model misspecifications, then to begin evaluating fit of the structural model given a well-fitting measurement model.

Anderson, J. C., & Gerbing, D. W. (1988). Structural equation modeling in practice: A review and recommended two-step approach. Psychological Bulletin, 103(3), 411–423. https://doi.org/10.1037/0033-2909.103.3.411

However, there are limitations with this method. A recent development improves on its limitations: The "Structural-After-Measurement" (SAM) approach to SEM. The PDF preprint of the Psychological Methods paper can be downloaded at that link.

https://osf.io/pekbm/

when reporting this, do I report the measurement parts that I obtained from step 1, or from step 2?

I would report the initial CFA's fit, discuss what was wrong with it and how you decided to fix it, then report the fit of the respecified CFA and its parameter estimates (e.g., loadings). The fit of the structural model is tested by comparing it to the CFA. If it fits as well as the CFA, then you can report its estimates (e.g., regressions). Posting on online supplementary appendix of all fitted models greatly enables transparency, facilitated by new repositories like the open-science framework.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.