Based on my understanding, online survey data is considered as convenience sampling. It is a non-probability sampling and not suitable for further analysis with SEM modelling.
However, some of my colleagues have a different point of view. I am confused.
Is there any article which I can study to get more knowledgeable about this?
There's been a recent exchange on this topic on the SEMNET listserv. You can search the SEMNET archives by registering with your email address (it's free). When you look at the discussions on 1/27/2023, you will find a thread on "Statistical models using non-probability samples" that addresses your concern.
If you want to be strict about such definitions, then accept that all questionnaire-based survey samples are "convenience" samples.
Even if you do manage to have a list, or other sampling frame, that captures the universe of your population of interest, then you only need one potential respondent to decide to not answer your questions and then immediately you no longer have a "random" sample.
The purpose of a "random sample" is to be confident that your sample is representative of your population of interest, with minimum risk that the sample could be biased in any way.
This is why most research reports in the social sciences include notes and tables explaining summary demographics - so that the reader can infer that the sample is acceptable for the research context.
- Badges
- Science method
- Similar topics
- Methodology
- Research Methodology
- Qualitative Research