I do find all of your videos very helpful. You explain very clearly. I use structural equation modelling using Smart-PLS, but have developed an interest in MPlus because of your videos. I look forward to a lot more sessions with you.
Hey :) I have like some independent latent variables which I hypothesize to all significantly influence a dependent latent variable. Furthermore I included one latent variable that acts as a moderator and I suggest that this moderator influences the correlation between the independent variables and the one dependent variable each in e negative way. Is it reasonable for this to just use regression ? in my mind it sounds quite simple to do all the ideas I have just mentioned with simple regression... Maybe you also have a recommendation for which program using ?
The primary advantage of SEM over regression is that SEM corrects for measurement error (unreliability). If your variables are highly reliable (if they contain little measurement error), using moderated regression (e.g., in SPSS or any general-purpose statistics software capable of regression analysis) may be fine. Otherwise, you might want to look into latent moderated SEM (Klein & Moosbrugger, 2000, link.springer.com/article/10.1007/BF02296338), which is implemented in Mplus. Best, Christian Geiser
@@QuantFish one further question when doing regression analysis. Do you know what’s the difference between OLS and Maximum likelihood estimation? I have read both in different papers as procedure for doing regression but I am not really sure what to use for my thesis…
@@ungeschult2168 The parameter estimates for a saturated linear regression model should be the same with OLS and ML. However, you may get slightly different standard errors, especially in smaller samples. See also: stats.stackexchange.com/questions/652138/why-different-results-for-linear-regression-in-sem-from-lavaan-and-lm Best, Christian Geiser
I do find all of your videos very helpful. You explain very clearly. I use structural equation modelling using Smart-PLS, but have developed an interest in MPlus because of your videos. I look forward to a lot more sessions with you.
I'm glad to hear that my videos are useful for you. Good luck with your work!
Best, Christian Geiser
Very cool. This is gonna help with my methods exam. Thank you.
Hey :) I have like some independent latent variables which I hypothesize to all significantly influence a dependent latent variable. Furthermore I included one latent variable that acts as a moderator and I suggest that this moderator influences the correlation between the independent variables and the one dependent variable each in e negative way. Is it reasonable for this to just use regression ? in my mind it sounds quite simple to do all the ideas I have just mentioned with simple regression... Maybe you also have a recommendation for which program using ?
The primary advantage of SEM over regression is that SEM corrects for measurement error (unreliability). If your variables are highly reliable (if they contain little measurement error), using moderated regression (e.g., in SPSS or any general-purpose statistics software capable of regression analysis) may be fine. Otherwise, you might want to look into latent moderated SEM (Klein & Moosbrugger, 2000, link.springer.com/article/10.1007/BF02296338), which is implemented in Mplus.
Best, Christian Geiser
@@QuantFish thank you very much !
@@QuantFish one further question when doing regression analysis. Do you know what’s the difference between OLS and Maximum likelihood estimation? I have read both in different papers as procedure for doing regression but I am not really sure what to use for my thesis…
@@ungeschult2168 The parameter estimates for a saturated linear regression model should be the same with OLS and ML. However, you may get slightly different standard errors, especially in smaller samples.
See also: stats.stackexchange.com/questions/652138/why-different-results-for-linear-regression-in-sem-from-lavaan-and-lm
Best, Christian Geiser
Thank you for your videos
Could you make a video about the Satorra-Bentler Chi-Square and how to compare and contrast?
Dear Tomas,
Thank you for the suggestion--will add it to my to-do list!
Best, Christian Geiser
Dear@@QuantFish Your videos are excellent, ... I will keep an eye out for your new releases, thank you.
@@tomasvargas-halabiphd2442 I'm so glad our videos are useful for you.
Best, Christian Geiser
Many thanks!!!@@QuantFish
Very helpful video! One question: Are the conditions for regression instead of SEM referred to in a handbook or article?
I'm not aware of a specific reference right now but many intro books on SEM likely cover this issue directly or indirectly.
Best, Christian Geiser
@@QuantFish Thank you very much Christian
How do we check reliability of the measurement during the planning phase of our study??
Sometimes, the previous literature can be helpful or you could conduct a pilot study.
Best, Christian Geiser