
Message/Author 

Abdel posted on Monday, March 03, 2008  6:34 am



Can a sample size be too large when using EFA? I am doing a EFA on a set of 10 categorical items. When I did a testrun on a sample of about 3000 subjects, the findings were exactly as I expected them and theoretically very interpretable. When I did the test again on a sample of about 10.000 subjects, the outcome was suddenly much less interpretable and the amount of factors that gave a nonsignificant chivalue with the small sample suddenly gave a significant chivalue. I used a sample of 10.000 subjects, because I plan to follow up the EFA with a CFA, which requires such a large sample to avoid empty cells. I was not able to find any literature on a possible maximum sample sizes. I hope someone can help me on this, thanks in advance! 


It sounds like your sample of 3,000 is not a random sample of the 10,000. I would think the differences are a function of the two samples having very different sample statistics and not the size of the sample. 

Abdel posted on Tuesday, March 04, 2008  6:00 am



The means and variances do not differ that much, but I did find a clear age difference between the two samples. Maybe that has something to do with it.. Thanks for your answer! 

Back to top 

