Lavaan package update discrepancy

[Taylor Fisher]

Hello everyone,

I am working on a project and noticed a discrepancy in lavaan outputs and wanted to reach out to see if anyone else had an issue with this and if they've found any guidance on it. Below you'll see I've built a latent factor and fit it using lavaan 0.6-12 and it shows the typical output with the standard and robust outputs (I'll be using the robust outputs).

However, when I run this same factor model in lavaan 0.6-13, I suddenly am met with standard and scaled results. 

My initial thought was something is changing the way it's choosing the estimator as I know it will assume based on the structure of the model and its components. However, I see no reason why that should have changed considering the estimator and optimization method have been reported and did not change. My assumption is that this is just a bug in the new version of lavaan and that the "scaled" results are the correct, robust output.

Has anyone seen this, or read anything about it to confirm that it is just a bug or language change and that the output has not been altered in some way? I want to confirm before I submit any models for publication.

Thanks,

Taylor Fisher

[Yves Rosseel]

It is not a bug. It is a change of 'labeling'. Instead of 'Robust',
lavaan (0.6-13) now prints 'Scaled'. Because that is a more accurate
description of what is going on: the test statistics (in the first
column) are literally scaled by a certain scaling factor (and sometimes
shifted) to obtain the results in the second column.

The term 'Robust' is rather vague, and may be too optimistic in some
settings. For example, the 'scaled' RMSEA or CFI fit measures turn out
to be not so robust after all. See for example

Savalei, V. (2021) Improving Fit Indices In SEM with categorical data.
Multivariate Behavioral Research, 56(3), 390-407.
https://doi.org/10.1080/00273171.2020.1717922

Yves.