What sample size considerations are critical for achieving robust results in SEM, especially when using complex models with multiple latent variables?
Here is a link to a widely cited article on sample size and SEM.
Article Sample Size Requirements for Structural Equation Models: An ...
Yes you should have 100-500 sample size for complex models. But sample size also depends on your research design and number of parameters . Best ratio is to have 20 on each parameter There is also a ratio of universe to sample size. To retain probabilistic distribution and representation. In no case more than 5500 samples are required even if it is pan country study . For T test etc you can use smaller sample size but not for SEM modelling . It all depends on accuracy you want and nature of study tools to be used and the size of universe. SEM and path analysis is trying to find multiple relations between number of parameters. Suppose parameters are 10 then size should be around 200 for best results.
Hi Densh! The sample size depends on several factors. First, the size of the population from which the sample will be extracted. That is because to generalize the results the sample must be representative of the population. Second, the SEM method to be applied. E.g. CB-SEM requieres higher samples for obtaining correct estimations of the parameters. To validate a CB-SEM model with 4 or 5 constructs we need a sample size of at least 250 observations. If the model has higher complexity the sample size should be much higher. Sample size also depends on the statistical power you want to achieve. In this case, a statistical analysis should be performed to obtain the sample size.
See Hair, J.F., Hult, G.T.M., Ringle, C.M., & Sarstedt, M. (2022). A Primer on Partial Least Squares Structural Equation Modelling. 3rd Ed. Thousand Oaks: Sage.
- Badges
- Science topic
- Similar topics
- Publishing