Dr Lin, you are an exceptional teacher. I have found many sources to explain CFA in a very inaccessible way. Here, you have explained CFA in an accessible way to those new to CFA, and made it free for all to benefit from. Thank you!
This is the resource I needed for my Dissertation. As someone with no strong stat and R background, this really helped me. Thank you very much Dr. Lin!
Thank you very much for an excellent lecture on CFA. Just a small comment/correction on the very final exercise: the Test statistic for the User Model, is 554.191. In your solution, it is 562.790. and the Degree of Freedom is 20, not 21. By putting these numbers in the formula, we get the correct CFI, which is 0.871 (rounded).
Wow, this was an absolute lifesaver. Such a complex topic, presented so simply and clearly, and freely available on TH-cam. I definitely owe you a beer. Thank you so much!
Thank you for this beautiful explanation! You have made my life so much easier! Quick question, do you think the poor fit (as indicated by the fit indices) was due to the fact that some items were not reverse-coded?
At 57:39 you say "make sure you are okay with setting the scale of that item". Lets say all items are measured on the same scale (likert 0 to 5). How do we assess whether we are okay with setting a particular item as the reference item? Spontaneously Im thinking that in the case of a likert scale where all items are on the same scale, we would atleast want to ensure that the item that is used as the reference item is an item where each answer option (0, 1, 2, 3, 4, 5) has been used "equally" often? For example, using an item where 4 was not chosen by a single respondent - seems like a bad item to use as a reference item? Is this at all true?
great video! Thanks. What do you think of using the estimation method DWLS instead of ML for ordinal items (such as those in the video "strongly disagree to strongly agree")? I have just read a paper (Reimann et al. 2024) where they used DWLS in a 2-factor CFA and got a great RMSEA (0.01). Their rationale was that the responses are ordinal and not continuous. Interestingly, I could run the same data set with ML and got an RMSEA = 0.13. Obviously a big difference. In papers, authors often do not even mention their estimation method.
Thank you for the great seminar. Would you please tell me where I can find materials about the following items:Two-item factor analysis Uncorrelated factor analysis with two items
Really great lecture! It helped me a lot with writing my thesis. One question though: If you use your full data, what is the default calculation running in the background? Does lavaan calculate a covariance matrix or a correlation matrix? Thank you :)
Thank you for this video. I followed the steps but I got an error message "covariance matrix of latent variables is not positive definite". There aren't any negative values in the covariance matrix though and also not in the correlation matrix. Grateful for any help to fix this issue.
Still can't believe this is free to watch for everyone. Thank you so much.
Dr Lin, you are an exceptional teacher. I have found many sources to explain CFA in a very inaccessible way. Here, you have explained CFA in an accessible way to those new to CFA, and made it free for all to benefit from. Thank you!
You're welcome Josh. Glad you enjoyed the content.
This is the resource I needed for my Dissertation. As someone with no strong stat and R background, this really helped me. Thank you very much Dr. Lin!
Thank you very much for an excellent lecture on CFA. Just a small comment/correction on the very final exercise: the Test statistic for the User Model, is 554.191. In your solution, it is 562.790. and the Degree of Freedom is 20, not 21. By putting these numbers in the formula, we get the correct CFI, which is 0.871 (rounded).
Wow, this was an absolute lifesaver. Such a complex topic, presented so simply and clearly, and freely available on TH-cam. I definitely owe you a beer. Thank you so much!
This is, hands down, the most accessible explanation of CFA I have seen/read. Thank you SO much!
You're welcome! Thanks for watching!
The most useful source I've found for my thesis, thank you!
Hei, I really wanted to say THANK YOU. This video really helped out with CFA. And I have never learnt about it before! You excel as an educator!
Best video on CFA ever. I understood almost everything. Thank you.
I am very grateful to you for clearly explaining with all the details. Thank you so much. Stay blessed!
These videos are fantastic. Thank you!
Thank you doctor Lin, really a simple explanation its very important…, you did it excellent!…
This is awesome, thank you!
great job mah man
In 2:03:19, he said "I am just showing you the real world", I died.
Thank you for this beautiful explanation! You have made my life so much easier! Quick question, do you think the poor fit (as indicated by the fit indices) was due to the fact that some items were not reverse-coded?
Dr. Lin, is there any video on your seminar on EFA?
At 57:39 you say "make sure you are okay with setting the scale of that item". Lets say all items are measured on the same scale (likert 0 to 5).
How do we assess whether we are okay with setting a particular item as the reference item? Spontaneously Im thinking that in the case of a likert scale where all items are on the same scale, we would atleast want to ensure that the item that is used as the reference item is an item where each answer option (0, 1, 2, 3, 4, 5) has been used "equally" often?
For example, using an item where 4 was not chosen by a single respondent - seems like a bad item to use as a reference item? Is this at all true?
great video! Thanks.
What do you think of using the estimation method DWLS instead of ML for ordinal items (such as those in the video "strongly disagree to strongly agree")? I have just read a paper (Reimann et al. 2024) where they used DWLS in a 2-factor CFA and got a great RMSEA (0.01). Their rationale was that the responses are ordinal and not continuous. Interestingly, I could run the same data set with ML and got an RMSEA = 0.13. Obviously a big difference. In papers, authors often do not even mention their estimation method.
Thank you for the great seminar. Would you please tell me where I can find materials about the following items:Two-item factor analysis
Uncorrelated factor analysis with two items
Hello, thank you for the video. Very insightful. Please, how can I get factor scores for the latent factors?
Is the advanced seminar available online?
Hello, Author. Could you tell me how to get the residual vairances of a MSE by lavaan()? Thanks
Really great lecture! It helped me a lot with writing my thesis. One question though: If you use your full data, what is the default calculation running in the background? Does lavaan calculate a covariance matrix or a correlation matrix? Thank you :)
Thank you for this video. I followed the steps but I got an error message "covariance matrix of latent variables is not positive definite". There aren't any negative values in the covariance matrix though and also not in the correlation matrix. Grateful for any help to fix this issue.
My cfi is 1 and rmse NA
My rmse is more than 0.8