Hi Erin! I have a confusion that here you are performing CFA for one construct...but if don't we have to perform cfa between other constructs(like the entire model)?
Dear Dr. Erin, I LOVE LOVE LOVE your videos!!! Forever heartfelt thanks to you!!! I have two questions after watching the video(s) (assignment video included).: 1) My first-order and second-order models are identical in results The bifactor model has poor loadings on some item. The question is should I go for the first-order or second-order model. Since you did not really explain how to decide and why in your assignment video, my guess is to go for the second-order model for 3 reasons: 1) if the theory goes for second-order, then it should be second order, 2) one of the latent loadings in the first-order model is .6, which is a bit too high 3) statistically, a simplified model is always preferred, so second-order model is better. 2) After CFA, I am to check the concurrent validity. Do you have any videos featuring that (esp. for a second-order model measurement)? Gratefully, Charlotte
The first and second will usually have the same results - the difference is that you are suggesting why the latent variables are correlated - otherwise the model structure is effectively the same (I assume you mean first and second order hierarchical here and not bifactor, as these are pretty different). I usually tell people to go for the second order to help explain high correlations between latents if it theoretically makes sense. For validity, usually you correlate the overall/sum/average scores with another gold standard measure.
Dear Dr. Erin, Thanks for the amazing video. I have a few questions. I'm working on a hierarchical model and the model has a fit issue. Would the model specification differ if the construct is formative instead of reflective? Plus, how can we fix the negative latent variance issue, particularly when the standard error isn't identified? Thank you for the help.
Check to make sure it’s coded properly. Then try setting the error variance to a small positive number. I think I have some examples of this in a different video.
Hi Dr. Erin (: --- can you do a hierarchal MGCFA in r? I have been looking for tutorials as well as articles about this topic in general-- but, I have found no leads. Do you have any recommendations for me ?
Depends on what part you want to MGCFA - if it's the Brown steps I go over in my lectures, then you could just add the hierarchical part and it will work the same.
Thank you, it is really helpfull. I have additional question. Let assume, i have first order factors that are related with second order and what if i also have additional items for second order factor?Should i add one more equation for my second order that contains only its related additional items?
can you help in solving the following problem, whenever i try to edit an output graph I got the following message "cant edited because it created with qgraph
![[Live] : ONE 169 วันนี้!! "อนาโตลี vs อูมาร์"](http://i.ytimg.com/vi/IK3YZVWuFJg/mqdefault.jpg)
awesome, thanks!
Hi Erin! I have a confusion that here you are performing CFA for one construct...but if don't we have to perform cfa between other constructs(like the entire model)?
Sorry, I don't understand the question.
Dear Dr. Erin,
I LOVE LOVE LOVE your videos!!! Forever heartfelt thanks to you!!!
I have two questions after watching the video(s) (assignment video included).:
1) My first-order and second-order models are identical in results The bifactor model has poor loadings on some item. The question is should I go for the first-order or second-order model. Since you did not really explain how to decide and why in your assignment video, my guess is to go for the second-order model for 3 reasons: 1) if the theory goes for second-order, then it should be second order, 2) one of the latent loadings in the first-order model is .6, which is a bit too high 3) statistically, a simplified model is always preferred, so second-order model is better.
2) After CFA, I am to check the concurrent validity. Do you have any videos featuring that (esp. for a second-order model measurement)?
Gratefully,
Charlotte
The first and second will usually have the same results - the difference is that you are suggesting why the latent variables are correlated - otherwise the model structure is effectively the same (I assume you mean first and second order hierarchical here and not bifactor, as these are pretty different). I usually tell people to go for the second order to help explain high correlations between latents if it theoretically makes sense.
For validity, usually you correlate the overall/sum/average scores with another gold standard measure.
THANK YOU, Dr. Erin. (Literally shouting)
Dear Dr. Erin, Thanks for the amazing video. I have a few questions. I'm working on a hierarchical model and the model has a fit issue. Would the model specification differ if the construct is formative instead of reflective? Plus, how can we fix the negative latent variance issue, particularly when the standard error isn't identified? Thank you for the help.
Check to make sure it’s coded properly. Then try setting the error variance to a small positive number. I think I have some examples of this in a different video.
Hi Dr. Erin (: --- can you do a hierarchal MGCFA in r? I have been looking for tutorials as well as articles about this topic in general-- but, I have found no leads. Do you have any recommendations for me ?
Depends on what part you want to MGCFA - if it's the Brown steps I go over in my lectures, then you could just add the hierarchical part and it will work the same.
Thank you, it is really helpfull. I have additional question. Let assume, i have first order factors that are related with second order and what if i also have additional items for second order factor?Should i add one more equation for my second order that contains only its related additional items?
Yes!
can you help in solving the following problem, whenever i try to edit an output graph I got the following message "cant edited because it created with qgraph
Sounds like you probably can't edit the graph because of the way it's constructed - what code are you running?
@@StatisticsofDOOMDOOM path diagram from the output selection menu on sem by jasp
@@jamalabu-alruz3746 Oh! I don't think you can edit those pictures from JASP.
If I ever see you I will kiss the ground you are walking on.
Thank you! This gave me a good laugh. :)