Latent growth curve modeling in R (January, 2020)
Вставка
- Опубліковано 17 чер 2024
- This video provides a walkthrough of a latent growth curve analysis using the 'lavaan' package in R. The presentation is based on an example provided by Schumacker & Lomax (2016; www.routledge.com/A-Beginners... ). A copy of the .RData file can be downloaded here (drive.google.com/open?id=1OSc.... You can obtain a copy of the text file used here (drive.google.com/open?id=1_WG..., and a copy of the Powerpoint I reference here (drive.google.com/open?id=1RJK....
Be sure to check out other videos on multilevel modeling (including growth curve modeling) here: sites.google.com/view/statist...
Thank you so much for your sharing!!!
Daaaaamn! I'm feeling it
Hi Mike. The links above appear to be not working. Would you mind refreshing them? Thank you.
Thank you very much for the valuable information you have provided. It is indeed very insightful. Could you also explain how to plot multiple graphs for continuous variables? Additionally, is it possible to add another variable that predicts the satisfaction variable and then repeat the study?
Your videos are always very helpful, Mike. Can you please tell me the best approach for two wave mediation model analysis? And whether you have already made a video on any method about two wave analysis already? Thanks.
Thank you so much for all these awesome videos, Dr Crowson! I was wondering if you could also make a video demonstrating LGCM where the growth model is non-linear? what can we do if the data do not fit a linear model, or how can we know if the data fit a linear or non-linear model using R??? Thanks a bunch for your help, really appreciate it
Thanks for this video. Unfortunately, the Google links do not seem to work anymore.
Dear Crowson, thanks for making such informative and easy-to learn videos. I have recently started learning multilevel mediation and moderation/moderated-mediation etc, and I found this subject very difficult to understand. Can you kindly make a video(s) on this topic as there is a great confusion and no one has elaborated the execution of these models such as 2-1-1 in simple ways as you usually do in your videos. I would request you to kindly make a video on this important topic and prepare videos for R & Mplus (both). Thanks.
Hi there, Qaiser. Believe it or not, I am currently working on materials to demonstrate how to do this using the 'lavaan' package R. I don't spend much time in MPLUS, but I will put it on my "to-do" list. But yes, I should have the 'lavaan' demo up and running in a fairly short while. Thanks for your request and for visiting my site!
@@mikecrowson2462 great to know all this. I'll passionately wait for the forthcoming video, in that case. About Mplus, I'll suggest to keep it for later time if there is any demand arise. In fact, everyone is shifting towards R thus it's better to concentrate more on it than something which is less looked for.
At the end, thank you again for your efforts & fingers crossed for the future videos.
Great video, thank you! Liked and subscribed. I find this example very useful in making quick distinction between indicators and regression in context of LGM. I have a question, though. I need to build a bivariate latent curve model, and I understand what it is from reading bollen and curran book p.199. Basically I have two series, y and w, y as you modeled here and another series w that I also want to model as indicators of a different growth process(different intercept and a different slope). My research question is actually concerned with how the latent slope of y and latent slope of w correlates with each other, or rather, "travel" with each other within individuals. Can I then simply specify two intercepts, two slopes. and in regression part regress one slope on the other? I haven't found an example in lavaan package tutorials itself, or any other sources...
My guess of the syntax would be like
model
Hi there. You look like you are going in the right direction. Keep in mind that when specifying a covariance you need to use double tilde (~~). For example, to specify a covariance between your factors, you'd need to use slope_y~~slope_w instead of slope_y~slope_w. (What you have specified from your syntax is a regression of the the slope_y factor onto the slope_w factor; in my example, I was using the single tilde for predicting variation in the slope and intercept factor from gender). To achieve a correlation between your slope and intercept you would use: slope_y~~intercept_y. And so forth...
So a single tilde (~) is specifying a regression where the variable on the left is regressed onto the variable on the right.
All that said, I don't think there is a need for you to specify correlated factors in your model. I'm operating from memory here (which may be faulty), but I believe lavaan will likely default by allowing a correlation between them. You can always test this speculation pretty easily by running your model with and without the correlation (using the ~~) to see if there is a difference. But I believe there won't be. I hope this is helpful to you.
@@mikecrowson2462 thank you so much! That's very helpful!!
How would you interpret the results, specifically the slope and intercept INTERCEPTS, if you had a continuous predictor? Like age? Would the intercept average be the value of your outcome variable when age is 0 or at mean age (if centred)? And the regression with intercept would be how much the outcome variable changes with every unit (year) increase in age at initial status?
Similarly, would the average slope represent the average change over time for individuals with mean age (if centred), and the regression value would be how much the slope changes with every additional year at initial status?
And how would your interpretation change if you were looking at standardised estimates in the output?
Hello
Thank you for your help.
Where is your codes for initial graph?
You are welcome. Actually, the figure I drew out was done in AMOS using the Growth curve model plugin. It was done a little differently than the authors' diagram, but same basic model. Unfortunately, I have yet to see a good drawing program in R for SEM. There are efforts, but no strong ones that I find yet. Best wishes!