Hi Christian, thank you so much for posting lots of useful videos! I am currently working on LGCM. I would like to use growth factors (measured in T1-3) as IV to Y (measured in T3) mediated by M (measured in T2). My problem is (1) whether it can be tested in Mplus and (2) if there is your video explaining the issue. Plus, we actually have 6 growth factors that are conceptually parallel IVs. Thank you again and it will be great if you could help us out.
Hi Dear. These Y are dependent construct right? Then what about the independent constructs? And what about those influence of these independent construct for example we if had moderator and mediation in time1 and time 2?
The Y variables represent the repeatedly measured observed variable (indicator) of the growth factors (e.g., IQ scores of the same people measured repeatedly over time). The intercept factor represents the true (measurement error-free) intelligence scores at Time 1. The slope factor represents true (error free) changes across time. You can add predictors, mediators, and/or outcomes of the growth factors to the model. You can also add time-varying covariates of the Y indicators. Best, Christian Geiser
Hi Raven, Yes it is possible but the items may have to be treated as ordinal variables if they have only a few response categories. Best, Christian Geiser
Why could this not be shown on an Y and X axel? With time being the X-axel, slope being the line/curve between the different points, intercept the starting point on the y-axel and epsilon makered by different colors?
It is indeed possible to plot individual and average growth curves in this way. See my other video here: ua-cam.com/video/0G-b5d8EoGg/v-deo.html Best, Christian Geiser
So the intercept represents inter-individual differences at the onset (time 1), and not at the other time points, right? Would it make sense to have several intercepts, each representing inter-individual differences at each time point? Or to have an Intercept representing inter-individual differences *across* all time points?
In a multiple-indicator growth model (e.g., second-order growth model, see ua-cam.com/video/5DXrvnjQ_6w/v-deo.html) you can have a factor at each time point that represents true (error-free) inter-individual differences at that time point. In latent-state trait models (see ua-cam.com/play/PL-kVjeOVYChquaZq6jrt0XDBYcOK3qkNa.html), you can specify an "overall" trait factor that represents the (stable) trait scores across all time points. Best, Christian Geiser
Dear Christian Geiser, thank you very much for the information and the quick response! I will look at your videos on these topics. Best, Paula @@QuantFish
@@QuantFish I have another question on multilevel cross-lagged models. Do you happen to have an example of multilevel cross-lagged models with several “random factors”/levels? In my analysis, measurements (2 variables measured at 4 time points) are nested in schools, and schools are nested in communes (so I have the two random factors school and commune), and I would like to test a cross-lagged relationship between the two variables. I am new to these models and MPlus, so I am still in the process of getting acquainted with the model types and code. Thank you again for your great videos! Best, Paula
Thank you for making statistics clear so I can truly understand what I am learning in class. You are wonderful!
Thank you very much for your very nice note. I really appreciate it. I'm glad my videos are helpful for you!
Best, Christian Geiser
Hi Christian, thank you so much for posting lots of useful videos! I am currently working on LGCM. I would like to use growth factors (measured in T1-3) as IV to Y (measured in T3) mediated by M (measured in T2). My problem is (1) whether it can be tested in Mplus and (2) if there is your video explaining the issue. Plus, we actually have 6 growth factors that are conceptually parallel IVs. Thank you again and it will be great if you could help us out.
vielen Dank
Hi Dear. These Y are dependent construct right? Then what about the independent constructs? And what about those influence of these independent construct for example we if had moderator and mediation in time1 and time 2?
The Y variables represent the repeatedly measured observed variable (indicator) of the growth factors (e.g., IQ scores of the same people measured repeatedly over time). The intercept factor represents the true (measurement error-free) intelligence scores at Time 1. The slope factor represents true (error free) changes across time.
You can add predictors, mediators, and/or outcomes of the growth factors to the model. You can also add time-varying covariates of the Y indicators.
Best, Christian Geiser
Hi Christian,
Thanks once again for this very helpful video!
Would it be possible to model single item variables via LGCM as well?
Hi Raven,
Yes it is possible but the items may have to be treated as ordinal variables if they have only a few response categories.
Best, Christian Geiser
Why could this not be shown on an Y and X axel? With time being the X-axel, slope being the line/curve between the different points, intercept the starting point on the y-axel and epsilon makered by different colors?
It is indeed possible to plot individual and average growth curves in this way. See my other video here:
ua-cam.com/video/0G-b5d8EoGg/v-deo.html
Best, Christian Geiser
So the intercept represents inter-individual differences at the onset (time 1), and not at the other time points, right?
Would it make sense to have several intercepts, each representing inter-individual differences at each time point? Or to have an Intercept representing inter-individual differences *across* all time points?
In a multiple-indicator growth model (e.g., second-order growth model, see ua-cam.com/video/5DXrvnjQ_6w/v-deo.html) you can have a factor at each time point that represents true (error-free) inter-individual differences at that time point. In latent-state trait models (see ua-cam.com/play/PL-kVjeOVYChquaZq6jrt0XDBYcOK3qkNa.html), you can specify an "overall" trait factor that represents the (stable) trait scores across all time points.
Best, Christian Geiser
Dear Christian Geiser,
thank you very much for the information and the quick response! I will look at your videos on these topics.
Best, Paula @@QuantFish
@@Paula-of5nf You're welcome. Good luck with your work.
Best, Christian Geiser
@@QuantFish I have another question on multilevel cross-lagged models. Do you happen to have an example of multilevel cross-lagged models with several “random factors”/levels? In my analysis, measurements (2 variables measured at 4 time points) are nested in schools, and schools are nested in communes (so I have the two random factors school and commune), and I would like to test a cross-lagged relationship between the two variables.
I am new to these models and MPlus, so I am still in the process of getting acquainted with the model types and code.
Thank you again for your great videos!
Best, Paula