R - Multigroup CFA Class Assignment 2

Lecturer: Dr. Erin M. Buchanan
Harrisburg University of Science and Technology
Fall 2019
This video updates the older version of the multigroup confirmatory factor analysis examples. This version uses a newer package and shows you how to complete the steps of a multigroup analysis even if you don't want to use that package. Learn how to compare groups in CFA for configural, metric, scalar, and strict residual invariance, followed by a few ways to calculate latent means. I also talk very briefly about equivalence testing - more on that to come!
A complete outline of the SEM course is here: statisticsofdo...
The materials for this lecture are on our OSF page: osf.io/2y67f/

[@wac37]

So by now I have been through the level 4 (out of 5) stuff in the major SEM books (except Brown). Gotta tell you, these videos are ridiculously good. I've coded along to 10 videos and I plan to use the code as templates for all my major SEM work in future. I think I'm gonna become my campus expert between your videos and the books, but your videos gave me an excellent start for overview and excellent finish for practical, step by step approach with model-validation at every step. Oh, and I almost forgot - fantastic interpretation of statistical output as well as amazing interpretation of the model (very useful in the write-up stage).
Every major SEM book I've seen, deals with only 3 out of the 5 topics I am interested in (MGCFA, Hierarchical, LGCM, MTMM & IRT). And you cover 90% of all those topics. Typically the major books add stuff like model-identification or math-equations or slightly different take on residuals or some nugget that takes you a bit deeper etc (like Beaujean's explanation of model identification from a simultaneous equation perspective (having enough info to solve the equation).

I can see why you're so extremely well published, everyone must want to work with you. Anywa, you gotta get that book on the market!! Or at least deliver a course on Coursera/Edx for a certificate.

[@StatisticsofDOOM]

Thanks for the kind words!

[@sedzinfo]

very nice, thank you

[@esra8205]

Hi Erin,
Thank you so much for all your effort! Your videos are excellent and my go-to for problem-solving.
I wanted to ask if you would consider the hypothesis of strict invariance as an omnibus hypothesis, of which the alpha error should be corrected in the context of multiple testing? I know it's quite challenging, especially since we're actually trying to verify the null hypothesis, so it's all a bit more complex anyway. It would just be a nice topic for the discussion part of my thesis.
Many thanks in advance!

[@StatisticsofDOOM]

This is an odd case that falls outside of NHST, so I would not do corrections, it's sort of already controlled for when using CFI or RMSEA.

[@Dobolinan]

Hi Erin I just completed your SEM course on datacamp and love your videos on the channel. I was wondering if the multigroup CFA approach can be applied to panel data to confirm that the structure between the latent and observed variables is invariant over time.

[@StatisticsofDOOM]

I would think you could apply these same basic ideas of multigroup structure to however you want to model the panel data ... are you thinking latent growth? That might have some special steps to it, as it already does some interesting things with slope/intercepts.

[@Dobolinan]

@@StatisticsofDOOM Thanks for the reply! I have not watched your latent growth model videos yet. Can a latent growth model tell me how the relationship between the latent variable and observed variables I model in a CFA might change over time? I have a CFA model using 9 indicators with multiple time points (23 time years). Can a latent growth curve model show how the factor loadings change over time?

[@StatisticsofDOOM]

Yes - might be fairly complex 😖 watch those first and then see if you think they will be what you want.