JASP - Structural Equation Modeling

[@EduNationOfficial]

Hi Erin... I'm from Indonesia... thank you so much for the video. My Student can learn Statistics with JASP now.

[@mountainsmusicbeer5532]

Brilliant. Thank you.

[@StatisticsofDOOM]

You're very welcome!

[@bucksman]

Thanks a lot for this video. It's life saver!

[@StatisticsofDOOM]

Glad to be of help!

[@TheAmandeep2009]

Keep doing the good work

[@Felisepu007]

Hi, Erin. I'm working on a CFA analysis using JASP. I was reading that for a 5 likert scale it is recommended to use a "Polychoric correlation" (or something like that?). Its that possible to perform in JASP???? I really appreciate your videos, they are awesome!!

[@StatisticsofDOOM]

I would suggest checking out this page that explains the idea between DWLS and polychoric correlations: stackoverflow.com/questions/41395611/cfa-in-r-lavaan-with-ordinal-data-polychoric-correlation-included So, lavaan can do it, but I am not 100% sure that JASP has all the options for it. I think if you click DWLS under estimator and then Mplus emulation, that should match it to ordered effects.

[@ergunkara1867]

many thanks

[@yazeedghanayim8909]

Hi, I want to thank you for your helpful videos, and I want to ask you whats the way to check mediation between latent variables in the JASP?

[@StatisticsofDOOM]

You’d have to code it the same way as mediation in lavaan - they have an example on their help pages.

[@zuhayrmustun4589]

Hello, nice video very informative! How do I check the validity and reliability of model? Thanks

[@StatisticsofDOOM]

For validity, you will need some outside criterion. I suppose you could do alpha for reliability but that would be of the scale not the whole model.

[@seongyonglee5747]

Hi Erin. I really appreciate all the videos you uploaded on KZbin. Now I wonder how I can get R^2 and t-values (boot trap?) from Jasp as we can from R programming. Thank you!!

[@StatisticsofDOOM]

At the moment, I don't see an option for R2 for lavaan - the JASP website has a place for suggestions though!

[@seongyonglee5747]

@@StatisticsofDOOM Thank you for your quick answer. Yeah. So, I posted a discussion about this issue in a JASP forum. I am now watching your videos on SEM with R. Thank you so much!

[@Cleciopesquisa]

Do you know where I find the explanation coefficient (R2)?

[@StatisticsofDOOM]

I do not see it at the moment :| I'll send them a message and see if they can add it.

[@marysmith2153]

Hi Erin. I really appreciate all your great videos! Big thank you :) Now, I am wondering if I should I activate the option "assume factors uncorrelated" or just leave it be when conducting a path analysis (social sciences) in SEM? I get very different outcomes depending on whether I activate this option. When it comes to comparing nested model, the degrees of freedom always remain the same and there is no significant improvement of the Chi Square when adding model paths. When this option is activated, I am able to compare my models. Do you have any recommendations on that option?

[@StatisticsofDOOM]

If it's path analysis - with only measured variables, often people do assume the "factors" uncorrelated - meaning here that it doesn't automatically add the exogenous only covariance to your model. I might suggest watching the path model example from my lavaan videos to help explain what is happening in the background when you check/uncheck that button. (kzbin.info/www/bejne/q5SVZZSNaa94ha8 around 10 minutes in).

[@marysmith2153]

@@StatisticsofDOOM Okay great, thanks a lot! :)

[@JudiantoTjahjoNugroho]

Can we do SEM with moderating var in JASP?

[@StatisticsofDOOM]

Yes, you should have full functionality, as you have to write the model code yourself. Check out lavaan.ugent.be/ to learn more about how to write that code.

[@jessperry1703]

Hi, can anyone advise me if latent moderated stuctural equation models can be analysed in JASP? Thanks.

[@StatisticsofDOOM]

I responded to your email but for others asking this question, you should be able to define: Y ~ X1 + X2 + X1:X2 for interactions.

[@alifarrukh9542]

is there way to create simulated data for SEM in R

[@StatisticsofDOOM]

Yes! Lavaan will actually allow you to create simulated data from a model - check out the lavaan guide rdrr.io/cran/lavaan/man/simulateData.html

[@Davao420]

omg , I shouldn't have paid for that AMOS license :-(

[@StatisticsofDOOM]

:( AMOS is nice sometimes, but you definitely get more control with lavaan - I was super happy to see them implement it in JASP!