3
$\begingroup$

I’ve run a pilot study with only 50 participants and I was wondering what I could actually do with a small sample like that, how can I best use this data before collecting the final sample of 600 participants. I will be doing an exploratory analysis and a confirmatory analysis. The goal is to validate a novel construct that is supposed to include 4 different sub-factors.

I did the EFA with the pilot study and it seems that the 4 factors formed a unique factor.

Are these preliminary results reliable, considering the small sample size? What else I could do?

$\endgroup$
2
  • 2
    $\begingroup$ That looks very small to get reliable results from CFA. Not sure why you want to do EFA first if you have a theory but people often do. $\endgroup$
    – mdewey
    Commented Feb 22, 2021 at 17:22
  • $\begingroup$ Sorry, I meant that I will do the CFA, and also the EFA, with the final sample (N=600). So I can consider the results from the EFA, of the pilot study (N= 50), reliable and so change my research design based on this results? $\endgroup$
    – Zuzanne
    Commented Feb 22, 2021 at 21:32

2 Answers 2

2
$\begingroup$

Unless you strike really lucky I think 50 is too small.

MacCallum, R. C., Widaman, K. F., Zhang, S. and Hong, S. in a paper entitled "Sample size in factor analysis" published in Psychological Methods, in 1999 volume 4, 84-99. doi:10.1037/1082-989X.4.1.84 discuss this in much detail as there is no simple answer.

$\endgroup$
1
$\begingroup$

The problem is that EFA (and CFA, kind of) is based on correlations. Look at the confidence intervals that you get on a correlation with N = 50. Here's a calculator: https://www.statskingdom.com/correlation-confidence-interval-calculator.html

A correlation of 0.3, with N = 50 has CIs from 0.02 to 0.53 - that's such a huge range that your correlations (and hence your loadings) will be all over the place - the same underlying population data could give rise to all sorts of different solutions.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.