Mplus Calculate Custom Estimates (CR and AVE) for validity and reliability
HTML-код
- Опубликовано: 21 авг 2024
- In Mplus, I show how to calculate custom estimates for composite reliability and average variance extracted. I also explain what all the syntax means. CORRECTION: at 4:42, I correctly say that we should take the square root of the AVE, but then I mistakenly proceed to square the AVE... oops!
Happy to see you back
Thank you for making this video!
Thank you James for sharing your knowledge, your videos are extremely helpful!!!
If possible can you kindly demonstrate Multitrait-multimethod correlation matrices (MTMM) for the evaluation of Convergent and Discriminant Validity (if possible for two different methods, such as Face-to-face interviewing and Telephone interviewing or in relation to time (test-retest stability, T1 and T2). Thank you in advance
I've never done MTMM, so I'm not sure. If I end up needing to conduct such an analysis, I will definitely make a video about it.
Thank you James for your quick reply. Much appreciated!
No worries for the MTMM. I will sort it out.
Thanks once again for sharing your extremely helpful videos!
Best wishes
Starting at 4:48, you squared AVE and you compared with correlations between each construct. However, the rule of thumb is just the opposite: AVE is compared with the SQUARED correlations between each construct (i.e. the square root of AVE is compared with correlations between each construct) (Reference: Fornell, C., & Larcker, D. F. (1981). Evaluating Structural Equation Models with Unobservable Variables and Measurement Error. Journal of Marketing Research, 18(1), 39-50.)
Did I misunderstand something?
Ha! Good catch! I totally messed that up. I even said square root, but then I squared it...
@@Gaskination please atone for your mistake by making a video about model comparison using CFA. In fact, the power of CFA over EFA is to check unidimensionality and compare different models. Normally in (top) journals, authors often compare their focal model to null model + uncorrelated factors + correlated factors + ...
A video exclusively demonstrates this is extremely helpful. You can comprehensively take into account fit indices and chi-square statistics (delta chi-square)...
I sincerely thank you if you make such a video.
@@ethiart This is a good suggestion. I'll add it to my list of videos to make. I'm not sure if I'll get to it super soon though...
@@Gaskination Patience is a virtue.
Hi James, thanks for putting up such a helpful video. I do have one question about the variance constraint on the latent factor though. Is there a reason you chose to set it to 1? Is that because the automatic Mplus constraint on the factor loading would be 1 or could one set it to any number?
That is the default value for covariance-based SEM. Also, here is an updated video with a macro to produce the tables for you: ruclips.net/video/O-PwcJDYVMs/видео.html
@@Gaskination thanks for explaining. The macro is great, but I have twolevel analysis, so the macro does't work my output. but having this way to compute CR in mplus is awesome :)
For calculating CR and AVE, is it necessary to include under MODEL the correlations among items we established when doing the CFA? Does it make any difference?
Mplus will automatically correlate all of the exogenous latent factors. So, it is not absolutely required to include "WITH" statements.
Sir If my factors indicators are 5 or 6 , so how can we named the error variances ?
is it okay like this :
OEA1-OEA4 (OEA5-OEA8);
LS1-LS5 (LS6-LS10);
TP1-TP6 (TP7-TP12);
Yes, that is correct.