I would agree with Manuel, there is no general reason why this would not work. In practice one needs to be able to make the argument that the sampling scheme has not created or minimized some structural link between concepts in a way that would not correspond to more comprehensive data. So the critical concept is sampling adequacy (including representativeness); adequacy may even be improved by careful sampling choices but there is nothing within a SEM that allows you to test this.

Good luck.

Siti -

If you are looking for "representativeness," as noted by Bruce, whatever the type of sampling, one of the best things you can do is to stratify well, placing data together that belong together, and sampling from each group (stratum).

For more on "representativeness," you might want to see what might apply to you from among the answers to the question at the link below.

Jim

https://www.researchgate.net/post/What_is_a_representative_sampling

how can i add  quota in sem mcp model in gams please help me