Obtaining a single unbiased SRMR value for a multi-group SEM

[Paul Connor]

I am planning to fit a multi-group SEM and if possible would like to follow Ximenez et al.'s (2022) recommendations for assessing model fit using unbiased SRMR values (USRMR) adjusted by communalities. lavResiduals() offers separate USRMR values for separate groups, but I am wondering if there is a way to get a single overall USRMR value for a multi-group SEM in lavaan, similar to how fitmeasures() returns a single SRMR value. I've tried googling this issue but haven't found any advice, so any help would be much appreciated.


Ximénez, C., Maydeu-Olivares, A., Shi, D., & Revuelta, J. (2022). Assessing cutoff values of SEM fit indices: Advantages of the unbiased SRMR index and its cutoff criterion based on communality. Structural Equation Modeling: A Multidisciplinary Journal, 29(3), 368-380.

[Terrence Jorgensen]

lavResiduals() offers separate USRMR values for separate groups, but I am wondering if there is a way to get a single overall USRMR value for a multi-group SEM in lavaan

Assuming you model the same variables in each group, you can simply average the (U)SRMR values across groups to get an overall measure.  But that only works for the point estimate, not its SE or CI.

Terrence D. Jorgensen

Assistant Professor, Methods and Statistics

Research Institute for Child Development and Education, the University of Amsterdam

http://www.uva.nl/profile/t.d.jorgensen

[Paul Connor]

Thanks Terrence. One question, though, does this also assume equal-sized groups? I anticipate some non-trivial differences in sample sizes.

[Terrence Jorgensen]

does this also assume equal-sized groups?

No, the things being averaged are elements in a (co)variance matrix. That’s why I said “assuming you are modeling the same variables in each group.”  The SE and CI would depend on sample sizes.