Hi,
I recently used a validated scale to measure self efficacy on technology use of maritime instructors. The scale was originally developed for measuring self efficacy for a generic instructor. One of the reviewer of my submitted manuscript insist that a CFA has to be done. Although my initial motivation was to just measure their self efficacy scores on technology use. Therefore, i wanted to know whether conducting a CFA is mandatory in such cases?
@Amit Sharma Since you have adopted validated scale(s) that were developed and used in other domain of research and you might have modified the items as per your study, it becomes the concern of construct validity viz. convergence and divergence. It is advisable to conduct CFA to establish the validity of your latent constructs.
You have to do invariance testing... so one model with a different number of samples, so you can test the samples.
Hello Amit,
You might be surprised as to how volatile a given scale can be when administered to a new/different sample in the same population, let alone when administered to a different population. I think that being able to say that the same factor structure given by others appears to hold for your sample would strengthen any arguments you make regarding your results.
Yes, it's a bit of a pain, but (the cfa) does help others to make judgments about how easy or difficult it may be to (slightly) repurpose the scale for a specific population.
Good luck with your revisions.
Amit Sharma, you have received some good advice up above. From my own experience, and from what I read in others' publications, different samples produce different factor structures even when the samples are overtly similar to each other and the scale appears to be equally relevant to people in those samples.
I think it's important to allow for the possibility that a CFA will not yield model fit with your data and that, therefore, using exploratory factor analysis EFA) would help you to identify what "went wrong".
Although it is not acceptable to conduct EFA and use those results to conduct CFA with data from the same sample (doing so is essentially a self-fulfilling prophesy), it is quite permissible to use the same data in going from CFA to EFA if you are puzzled by the initial CFA results.
All the best with your research - even if it means a degree of conceptual shuffling.
Simply put, if the context differs from the original scale’s, or any adjustments were made on its original version, a CFA could be required provided that proper sample size is used. You could refer to Schmitt (2011) for relevant methodological considerations.
Schmitt, T. A. (2011). Current methodological considerations in exploratory and confirmatory factor analysis. Journal of Psychoeducational Assessment, 29(4), 304–321. https://doi.org/10.1177/0734282911406653
Good luck,
Performing CFA is often recommended once you are using an instrument (validated or not) in a new study context, and population. Every population largely produces a different factor structure of an instrument, hence the need for CFA.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics