Hi
Is there any justification available in the literature for the minimum sample size for factor analysis (EFA and CFA), if the sample size is kept around 50 responses? the data have been collected from top officials of organisations, as they are limited in number that's why the data is limited.
The minimum sample size necessary for factor analysis depends on many different factors such as number and quality of indicators, the number of factors, magnitude of loadings, etc. Therefore, general recommendations cannot be made, other than that the number of free parameters to be estimated cannot be larger than the sample size.
If you want to make sure that your sample size is sufficient for stable and unbiased parameter estimates, the best course of action is probably to run a Monte Carlo simulation with the expected model and it's parameters and the given sample size. The simulation allows you to study whether N = 50 is sufficient in your case, given your specific model and parameters. For a free mini-course on how to run simulations to determine sample size requirements in factor analysis and structural equation modeling using Mplus, see goquantfish.com/courses/sample-size-planning-in-mplus
One commonly used recommendation is to have at least 5 observations per parameter. With 50 participants, that would imply a factor analysis with 10 or fewer variables.
FYI I personally think this is a simplistic criterion, but it is at least a starting point before you engage in complex simulations.
One of the key sources to answer this questions is: Article Sample Size in Factor Analysis
50 will be too low for some purposes, but as mentioned by Christian Geiser and David L Morgan it depends on several things. The more complex the model the more people are needed.
(1) For rotation purposes: at least 30 respondents. based on the context, 50 is ok to see the impact of your study.
(2) For the larger study, PhD/MSc level can go for either per variable (minimum 10 respondents) or per item (minimum 5 respondents). See the references as advised by Anastasia Kitiashvili.
(3) This is in line with the quantitative/survey method. If your context is associated with respondents at a higher level of the company, better follow qualitative approach (in-depth interview: Ref. Marshal 1996) with 15-25 respondents.
I do agree with Dr. Anastasia suggestions that has been given earlier to the solution. In addition, to the suggestions the researcher must justify why those respondents are sufficient for analysis.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size