Kim Fleche Since you have clear a priori hypotheses about the number of factors and how they are measured (i.e., with which items), I would use only CFA. EFA is most useful when you are exploring the number of factors and the loading patterns without a lot of prior knowledge/assumptions, which is not the case here.

Competing theories about cross-loadings of individual items can be more effectively tested within the CFA framework. Specifically, you can run competing models with different theoretical cross-loading specifications and compare the fit of such competing models, for example by using nested model likelihood ratio chi-square difference tests or information criteria (e.g., BIC).

This answer was very helpful, thank you for sharing!

Firstly, I'd like to say that this is a very interesting topic to research and it is very relevant these days. Unfortunately, donations are needed today more than ever to sustain the life of many people across the world. Furthermore, traceability was always an issue when it comes to decide on donations (whether to donate or not, if yes how much then through who etc..). Blockchain is one of those technologies that exist today to operate in environments where trust is not mandatory. I really wish you all the best with this! (by the way, I am currently working on the technical development of novel blockchains and also have good exposure on quantitative research, so perhaps we could collaborate on something?)

Regarding your question, I normally always conduct EFA prior to proceeding with CFA. When you use scales previously developed as part of separate models and combine them in a new one there is always a risk of cross-loading which can be very problematic in terms of model fit, reliability and explainability. EFA can help to ensure your measurement model is fit before proceeding to the structural model. Furthermore, EFA allows you to explore the possibility of creating new factors that were not necessarily thought of or discussed in the literature when you designed your theoretical model. Lastly, conducting EFA is quite systematic, not very hard to do and yet very useful, plus many reviewers expect it in studies employing SEM. However, always remember, it is not a race for the best model fit scores! Whether it is with EFA or CFA, keep the theoretical soundness be your guide even if this means some compromise on analysis results. A model that is theoretically sound with a goodness of fit of 0.95 is much more useful than a model we can't ground in theory with a score of 0.99.

Karim Al-Yafi Dear dr. Al-Yafi, thank you for sharing your valuable insights. Also, your final mention on (from my understanding) overfitting models is interesting (i.e., better fit is not always best model). Still, it seems like the approach that is recommended is different between various professors (see dr. Christian Geiser ’s thoughts above), which is always interesting. My research is merely for a master thesis in Marketing, so my understanding of the blockchain is quite basic and doesn’t cover knowledge on development of the specific technology. Nonetheless, I wish you good luck with your research!

Kim Fleche, I'm concerned that what seems to be your intention to conduct factor analysis (whether EFA or CFA) when you have scales with only two or four items might be akin to trying to kill a spider with a machine gun.

I wonder whether looking carefully at interitem correlations within each scale might be a more appropriate strategy.

That said, I think it might be useful to place all of the 10 items into an EFA to see whether there is undesirable overlap in the constructs that the three scales are intended to represent - something that I think you raised for good reason.

Apart from that, in my experience it's quite possible to get very nice goodness-of-fit results with CFA despite feeding near-nonsense into the analysis as long as there are at least three factors and a small number of items on each factor.

Every good wish for your research.

Robert Trevethan thank you for sharing your insights dr. Trevethan! I must note that my research consists of a few more constructs, but these consist of 2-4 items as well, meaning your remark is still applicable.

Your answer also gives me an understanding/explanation of why I found published studies that look at interitem correlations (reliability, most often with a Chronbachs alpha) and don’t seem to pay any attention to the validity of constructs by means of any type of factor analysis.

I agree, that it would add value to my study to perform an EFA anyway. Essentially so I can look if any items exist that crossload above certain treshold (and delete them) since this could be problematic if ignored.

Thank you once more!

Kim Fleche, I'm glad my post was helpful. Please be aware that coefficient alpha (often less appropriately referred to as Cronbach's alpha) is a much more complex, uninformative, and deceptive metric than many researchers seem to appreciate.

A major feature of coefficient alpha is that it is highly dependent on the number of items involved. Because of that, a small number of nicely correlated items can have quite a low coefficient alpha. Conversely, by the time there are 20 or more items, the value of alpha can be quite high despite little association between many of those items.

The following publications might be helpful:

Cho, E. (2016). Making reliability reliable: A systematic approach to reliability coefficients. Organizational Research Methods, 19(4), 651–682. https://doi.org/10.1177/1094428116656239

Cho, E., & Kim, S. (2015). Cronbach’s coefficient alpha: Well known but poorly understood. Organizational Research Methods, 18(2), 207–230. https://doi.org/10.1177/ 1094428114555994

Cortina, J. M. (1993). What is coefficient alpha? An examination of theory and applications. Journal of Applied Psychology, 78(1), 98–104. https://doi.org/10.1037 /0021-9010.78.1.98

Hoekstra, R. Vugteveen, J. Warrens M. J., & Kruyen, P. M. (2019) An empirical analysis of alleged misunderstandings of coefficient alpha. International Journal of Social Research Methodology, 22, 351–364. https://doi.org/10.1080/13645579 .2018.1547523

Schmitt, N. (1996). Uses and abuses of coefficient alpha. Psychological Assessment, 8(4), 350–353. https://doi.org/10.1037//1040-3590.8.4.350

Sijtsma, K. (2009). On the use, the misuse, and the very limited usefulness of Cronbach’s alpha. Psychometrika, 74(1), 107–120. https://doi.org/10.1007/s11336-008-9101-0 [There are commentaries additional to this article in the same issue of the journal.]

Taber, K. S. (2018). The use of Cronbach’s alpha when developing and reporting research instruments in science education. Research in Science Education, 48(6), 1273–1296. https://doi.org/10.1007/s11165-016-9602-2

Robert Trevethan thank you for such a comprehensive list of literature!

Robert Trevethan I agree with you that EFA can sometimes be helpful when you otherwise (with CFA) cannot easily figure out what's going on with the measures (e.g., which indicators cross-load on different factors). I disagree, however, that having "only" between two and four indicators per factor is necessarily problematic. Clearly, having more good indicators rather than fewer good indicators is preferable. However, in my experience, if you have the choice between fewer good (i.e., valid, reliable, and homogeneous) indicators versus more poor (invalid, unreliable, and/or heterogeneous) indicators, it is better to opt for fewer good indicators. CFA can even handle single-indicator approaches (see reference below), and in some cases those are preferable to multiple-indicator approaches.

Article Should researchers use single indicators, best indicators, o...

Christian Geiser, I suspect we are referring to somewhat different things in our posts. Nevertheless, in terms of a few good indicators versus a greater number of poor indicators, I couldn't agree with you more: quality trumps quantity.

Thanks for making me aware of the article concerning this. I've begun reading it with keen interest.