My psychometric scale contains 38 items in total, with 4 subscales (20, 6, 6 & 6 items respectively? Can i run a factor analysis for each subscale rather than the scale as a whole for the purpose of my small sample size? (approx 30 at the moment)
Hello Chantelle,
In my opinion, the purpose of factor analysis is, shortly, to analyze the internal structure of a test. Your questionnaire has 38 items that are theoretically grouped into four factors; In this case, you should carry out CFA and AFE to check that the items are grouped as you expect. It does not make much sense to carry out these analyzes individually for each subscale.
In any case I think you should reconsider increasing the size of your sample if the objective is to validate your instrument.
Best regards,
Javier.
Chantelle Fraser , can you say WHY you want to do factor analysis? It is unclear to me why you are trying to do these analyses. Do you just want to see if the items within a subtest are associated more with each other than with items not in that subtest?
Chantelle Fraser, given a related question you created for another thread, I think you are very much clutching at straws, particularly given your very small sample size and the kind of analyses you are wondering whether you might be able to conduct.
Sorry to deliver bad news!
In some studies, I also see that factor analysis was applied to the sub-dimensions of the scale. However, this is not the right method for me. Because, the purpose of factor analysis is to express the scale with fewer variables, instead of 38 items, if I will say for your study. If you apply explanatory factor analysis for each sub-dimension, you are doing something against the spirit of factor analysis.
Even if you did, it would only measure convergence and not discrimination so it is not going to tell you if your factors are covariant. Secondly, with n=30 I wouldnt trust the results as generalisable anyway. A robust approach would be to have n at least 150-200 and CFA all of the factors. Or just rely on other researchers’ validation of this scale (in its entirety).
Yes you can and you need strong references to support that, given the fact that the rationale of factor analysis is to follow the logic of the factor groupings. You refer to the paper below as reference source.
Amos, D., Au-Yong, C.P. and Musa, Z.N. (2020), "Developing key performance indicators for hospital facilities management services: a developing country perspective", Engineering, Construction and Architectural Management, Vol. 27 No. 9, pp. 2715-2735. https://doi.org/10.1108/ECAM-11-2019-0642
When you run sample size adequacy for factor analysis e.g. KMO, I don t think you sample will be suitable for factor analysis with n=30.
- Badges
- Science topic
- Similar topics
- Psychology
- Psychometrics