Our research team is developing a scale.
We have conducted EFA with a sample of 798 Chinese university students.
And since there are 34 items in the scale, we plan to conduct CFA with a different sample of 500 Chinese people. (at least 10 participants for each scale item)
I have two questions:
1. Is it OK that the sample in CFA is smaller than EFA?
2. Is it OK if the participants in CFA are not Chinese university students?
You can split the students sample into two subsamples. First run an EFA on half of the sample followed a CFA on the other half. You can also run another CFA (including the model you derived from the CFA in the student sample) on the other sample with 500 participants to further examine your CFA model. Then, you could run Measurement Invariance analysis across the two CFA samples.
Best,
Mojtaba
Is always advisable to use larger sample for your CFA, more than the one used for EFA.
I think you can use different Ns for EFA and CFA provided that each N is adequate for the analysis it's used in. Other than meeting the minimum requirements for each e.g. ratio of cases to items, there are no rules I know of concerning which one should be larger.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size