For an experimental design ( one experiment and one control group), at least or minimum, how many subjects need to be included in the groups in SEM?
This question depends on many factors including the structure of the structural equation model (SEM) to be specified, the number and reliabilities of the factor indicators, the effect size(s), and other data characteristics. Given then complexity of the analysis and the many factors that play a role, the optimal sample size for an SEM is most effectively studied through Monte Carlo simulations.
In my Youtube tutorial video below, I show how you can address the sample size planning issue for SEMs by running a Monte Carlo simulation in the Mplus software:
https://www.youtube.com/watch?v=oF_8ssvhmVg
I address the same issue in more detail in a free mini-course that is offered through QuantFish:
https://www.goquantfish.com/courses/sample-size-planning-in-mplus
See also:
Muthén, L. K., & Muthén, B. O. (2002). How to use a Monte Carlo study to decide on sample size and determine power. Structural equation modeling, 9(4), 599-620.
https://www.statmodel.com/bmuthen/ED231e/RelatedArticles/Article_097.pdf
Thank you for your answer Dr. Geiser, I have searched to find an article applying the experimental method using SEM. Can you suggest a good example of an article in social sciences?
Using Monte Carlo analysis may well represent "best practices" with regard to specifying the N in SEM, but the most commonly used approach is to have at least 10 observations per parameter in your model.
- Badges
- Science topic
- Similar topics
- Indexes