SEM with composites: composite variance parameters question in lavaan v.0.6-20

15 views
Skip to first unread message

Janine

unread,
Mar 25, 2026, 3:26:58 PM (10 days ago) Mar 25
to lavaan
Dear lavaan users, I have a few questions regarding the recent arXiv paper on SEM modeling with composites in lavaan.

In the Supplementary Material, under the "Variances" output, the composite variables (Quality, Performance, CorpSocResp, Attractiveness) do not have standard errors, and the lavaan output does not put a period in front of the construct names, which seems to indicate that these are "total" variances not residual variances. See below for screenshot. 

 

My questions are:

  1. How are these variance parameters calculated for the composite variables (Quality, Performance, CorpSocResp, Attractiveness)?
  2. Are these specified as causal-formative or composite-formative?

A.    If the former, why is there no estimated residual variance?

B.     If the latter, why are these values non-zero (given that residual variances in composite models are typically fixed to zero)?


Any guidance or insight would be greatly appreciated. 

 

Many thanks,

Janine


Supplementary Material Output Screenshot: 
Variance_Output.jpg

CitationSchamberger, T., Schuberth, F., Henseler, J., & Rosseel, Y. (2025). Structural Equation Modeling with Latent Variables and Composites. arXiv preprint arXiv:2508.06112.
Reply all
Reply to author
Forward
0 new messages