I have been looking for an interpretive explanation of bifactor's run in lavann which also assumes the audience has an intermediate understand (thus a focus on application). Not only did you deliver, the explanation is concise & precise, yet thoroughly covers common obsticals in a manner that is easily understood. Thank you for your public service!
Your video is so well explained and helped me a lot to understand it. Thank you very much! Such a good pace, you show the code and, mostly important, show it with an available dataset. I am trying to do a bifactor analysis with correlated factors and compared it with a 4-factor solution. Didn't work out, it says "covariance matrix of latent variable is not positive definite". Now I try to fix that. Greetings from Switzerland from a German student: Weiter so!
Thanks so much for dealing with identification issues, which are very common in bifactor models. One question. I see you removed the Visual factor, which was causing out-of-bound values. Do you think, instead, correlating the residuals of x1 and x2 and x3 after removing the Visual factor would have been an appropriate solution as well? Typically, residual correlations are not recommended but there is an exception: when you expect them to be correlated (forming a mini factor of some sort). Let's say x1, x2 and x3 are are all correlated moderately, would it be wise to specify the new model with these correlated residuals?
Interesting idea, one could try that, I guess. Since those items are theoretically connected there would be a possible justification for correlated residuals.
Thank you very much! Would it be possible to use the general factor for further calculations? E.g. to calculate the correlation between the general factor and another variable that was originally not part of the model? thanks again!
I have been looking for an interpretive explanation of bifactor's run in lavann which also assumes the audience has an intermediate understand (thus a focus on application). Not only did you deliver, the explanation is concise & precise, yet thoroughly covers common obsticals in a manner that is easily understood. Thank you for your public service!
Your video is so well explained and helped me a lot to understand it. Thank you very much! Such a good pace, you show the code and, mostly important, show it with an available dataset.
I am trying to do a bifactor analysis with correlated factors and compared it with a 4-factor solution. Didn't work out, it says "covariance matrix of latent variable is not positive definite". Now I try to fix that.
Greetings from Switzerland from a German student: Weiter so!
Thanks so much for dealing with identification issues, which are very common in bifactor models. One question. I see you removed the Visual factor, which was causing out-of-bound values. Do you think, instead, correlating the residuals of x1 and x2 and x3 after removing the Visual factor would have been an appropriate solution as well? Typically, residual correlations are not recommended but there is an exception: when you expect them to be correlated (forming a mini factor of some sort). Let's say x1, x2 and x3 are are all correlated moderately, would it be wise to specify the new model with these correlated residuals?
Interesting idea, one could try that, I guess. Since those items are theoretically connected there would be a possible justification for correlated residuals.
Thank you very much! Would it be possible to use the general factor for further calculations? E.g. to calculate the correlation between the general factor and another variable that was originally not part of the model? thanks again!
I think it is possible to correlate a general factor with a variable outside of the construct, yes.
@@RegorzStatistik cool! vielen Dank!