R - Multigroup CFA Lecture

[@sabo2725]

Thank you for this lecture. It really helped so far understanding the concept of multigroup CFA. At 13:16 you mention, when testing the groups separately with the same CFA structure, if it it seem to work for one but not the other group you should start exploring why the model fit indices are bad for one and not the other.. I was wondering how you would approach exploring the differences at that stage?

[@StatisticsofDOOM]

I'd start with modification indices to see where some of the bad paths or incorrect specification might be.

[@agathabataille3008]

Hi, Thank you so much for this lecture, it is incredibly useful. I have been trying to conduct it on my own data. However, the sample size of my group have less observations than the number of parameter with results in the non-positive definite warning. Any advice on what I should do? Can I still proceed with the study or should proceed differently? To give some context, I am testing multiple scales simultaneously as I am interested to see if the SEM model (with mediation) that I am testing works for both groups I am studying.

[@StatisticsofDOOM]

You definitely need to resolve the error - models with more parameters than observations are not valid.

[@Mike-ep8yl]

Thank you so much for the video. Can you direct me to the R code to test multi-group structural invariance? In other words, after confirming measurement invariance of my model, I would like to test whether the regression coefficients between latent factors are significantly different by group. I have been looking for the code in other videos you made and on other websites with no luck.

[@StatisticsofDOOM]

I have a bunch of examples of MGCFA - that would be the metric step in this video and you can find examples of code at statisticsofdoom.com/page/structural-equation-modeling/

[@paulariasmedina1927]

1:05:51 I can´t make it work and it´s driving me nuts... in partial_syntax I replace "~~" for "=~" (because I stopped at metric invariance, delta CFI >.01), then in object temp I wrote group.equal = c("loadings"). Everything else is the same. All my CFI are the same, I even tried manually testing one model at the time and still I get the same CFI for all the models... what am I missing.

[@StatisticsofDOOM]

You changed it to =~ right? Otherwise, it may happen that nothing will make the model better (I've had this happen). I would check out the loadings and see if they are very different.