Cross-lagged panel models

[@hisparussky]

Wow, man.. thank you sooooo much!!! Reading from a book, article or blog is completely different than hearing and getting the info explained by someone. Thank you a lot!!!

[@mronkko]

You are welcome!

[@drryangagnon]

Excellent work! Very clean explanation and helping me to consider alternative models with my future work. Thanks for putting this online!

[@mronkko]

You are welcome!

[@susi9240]

Thanks a lot for this awesome explanation! It helped me a lot with my thesis and saved me a lot with my help!

[@mronkko]

You are welcome.

[@gabrielarantestiraboschi2218]

Hi Thank you so much for this video! Great work! There is only one thing that is confusing me, and maybe you could kindly clarify to me: you mention that the correlation between random intercepts estimates the within-effects. But the idea that I got from reading the Hamaker paper (and from other sources) is that the correlation between the RI is actually the between-effect. What's the correct interpretation?

[@gabrielarantestiraboschi2218]

I noticed that later you mention that in the RI-CLPM the correlation of the RI is the between effect. So I am still a bit confused. Could you please clarify? Thank you.

[@mronkko]

@@gabrielarantestiraboschi2218 The correlation between the RIs the between effect. Can you give me the timestamp of the part of the video where I say otherwise and I can check it.

[@gabrielarantestiraboschi2218]

Thank you for your response @@mronkko. It is at 4:49 . Once again, great video and thanks!

[@nico.teutelbier8229]

Thank you very much for your explanation! I have the problem that I have 4 variables for my model with two measurement points each and therefore my model is not identifiable/saturated if I want to calculate the cross - lagged paths, the autocorrelations as well as the synchronous correlations. How can I solve this?

[@louistay6424]

Thanks for your full explanation! I have the problem that how to correct concomitant variables in this model by MPLUS software.

[@mronkko]

Thanks for the thanks. I would need more information on your mPlus problem to give any useful answer. If you want to control for a possible confounder, I need to know if it is a level 1 or level 2 variable and whether you are using the CLPM or RI-CLPM model.

[@louistay6424]

@@mronkko I greatly appreciate for your response.There are 4 level 1 variables to be corrected in CLPM model.

[@mronkko]

@@louistay6424 You would add them as exogenous predictors to "within" part of the model. In the typical CLPM model, you have two variables but if you want control variables, you will just add more variables. Just don't specify any predictors for these variables and make sure to allow them to be correlated with the errors terms of the main variables that are used as predictors along with the controls. For example Mplus syntax for CLPM, see: www.researchgate.net/publication/328095575_How_to_run_a_multiple_indicator_RI-CLPM_in_Mplus

[@DrStro]

Great job, thank you. I have one question regarding acceptable variables scales. I always assumed both variables should be continuous in order to perform CLPM, but in your first example (basic concept), you use "CEO gender" which lead me to the question - is there a way to use CLPM for two variables - one continuous and one dichotomous?

[@mronkko]

Yes. The nature fo the variable distributions is generally not very important for variable distributions. As long as you can assume linearity, things are simple. It is a common misconception that noncontinuous variables would require special procedures because of their noncontinuous nature journals.sagepub.com/doi/full/10.1177/1094428121991907

[@Nbl.369]

What if we have sequential mediation model in first point Time . That would definitely be one_ IV 2 _mediator and one _Dv If we have two times data with same participants with six months time interval. How would i analyze longitudinal study of two waves? Please guide

[@mronkko]

You can estimate models with two-wave only if you assume no unobserved heterogeneity. Whether that assumption is reasonable depends on your research question and context.

[@farahdeeba9417]

thank you very much. You are awesome!!

[@mronkko]

You are welcom!

[@jounlee5147]

Would RI-CLPM work with binary variables?

[@mronkko]

Yes. There is no problem with using linear models with binary data.

[@jounlee5147]

@@mronkko Thank you for your response. Do you have experience with MPlus?

[@khalilelbachiri9977]

Hello sir, I would like to ask you a question: I come across the term persistent process a lot but I can't find an exact definition of this term. Could you please give us a definition of the latter if possible?

[@mronkko]

I am not familiar with that term and could not find it in any of the articles or books that I have as PDFs. More generally, a term can refer to multiple concepts and which concept is referred to depends on the context. Moreover, concepts be defined in multiple different ways (see Jaccard and Jacoby, 2021 book on this) so even if I was able to give you a definition, it might not be the same definition that the authors who you saw use the term used.

[@khalilelbachiri9977]

@@mronkko thank you very much professor