I have six variables in my model and I want to test the relationship between them. my questionnaire contains scales of 5 variables along with one scale that I am developing. for my pilot study, after collecting data, I have conducted EFA and CFA. Do I need to conduct SEM to show how fit my model is at this stage? or SEM should be done after collecting the final data?
It depends on how many cases you collected and analyzed for your pilot testing. Commonly, the pilot testing does not require a rigorous statistical rigor based on a sufficiency of the sample size. I think that EFA and CFA are enough to send your paper to the journal if the journal accepts the pilot study. However, when you report the final research findings, SEM is strongly recommended to ensure the scientific rigor and justification for your argument.
Claire Su-Yeon Park
thank you for your answer. for my pilot study I have collected 54 data points. since the sample for my questionnaire is rare, the small sample size for pilot is adequate as reflected by the KMO value. i have conducted both EFA and CFA and I am going to use SEM after final data collection. I wanted to confirm if I required SEM for pilot test as well
Based on my opinion, during a pilot test, I will test the reliability of the instrument scales (Cronbach’s alpha) because I will collect only 30-45 cases. In addition to, small sample are not enough for running EFA, CFA, model fit, SEM. Pilot test aims to assess quality of questionnaire and you may find some unclear items which need to revise in final questionnaire. If you need to improve some items in final questionnaire, the results of EFA, CFA, model fit, or SEM which are run during pilot test may be not useful.
As reliability scores often depend on sample size , pilot studies (with 30_50) observations may give you very low reliability values. The pilot should therefore be used as a more general report of instrument quality.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling