Using technology acceptance models like UTAUT and DeLone & McLean/Wixom Todd, I am trying to develop a framework for Social Construction of Digital Library. I have added 3 variables in this new factor structure. Can CFA alone will be sufficient to check its validity? As EFA is practically not possible with a large number of items in the questionnaire. Also, as it is said that while adapting existing scales EFA is not necessary. But in this case, I have added a few variables to the existing scales. Please suggest and provide references.
Experiments regarding the measurement of time might be a good place to have a look. Each leap in precision required new measurement approaches.
You can use the two scales in developing your scale. However, you will still want to test the reliability and validity of your new scale. When you have a scale that has some but not all of other scales, your new scale may behave in different ways than the original.
We did just this when we developed the HeartQol scale. I've linked it below. You might well consider the methodology for your project.
Article The HeartQoL: Part I. Development of a new core health-relat...
I believe that CFA is more than adequate if you are testing the hypothesis that the three new items can be included in the original factor. In other words, you would test a model with two correlated factors versus a one factor model.
since you have an explicit hypothesis, I don't see any reason to run an EFA.
Agree with others. CFA should be adequate. Speaking of Likert items, the new items are supposed to load to one of the existing scales. Whether their loading is adequate can be assessed with a single factor FA for each scale with the new items. Then correlate the scales - if you want to combine them then a relatively high correlation should be established, e.g. >.85. If this is achieved, then you can proceed with selecting best items from each of these FA's and construct a new scale, assessing its reliability and validity again. If the reliability is > 0.7 and the single-factor solution is attainable, you can use the new scale in further analyses.
Don't know about the context of your research or the scales you are combining, but if you are taking items from established scales, adding some items, and calling it a "new" scale, some reviewers may not respond favorably to that, especially if they are the authors of the scales! A researcher once did that with 22 of 28 items from a scale I developed, and I was not particularly pleased that he represented the scale as a "new" measure without mentioning that all of the items were from existing measures. Another thing to consider is that many (most?) reviewers for journals focused on measurement and psychometrics would probably not consider CFA to be a very rigorous method for validation. You might take a look at the validation methods used in articles published in a journal that you might submit to.
Thanks so much David, Nikolaos, Eve and Jan. The insights given by you are very useful to me. Eve, I am mentioning in my research that I have derived the 56 items from three existing scales (on technology Acceptance model-TAM) and adding my own 7 items for developing a model. I am not calling it a scale as I am not validating it through EFA. I am just calling it a model to be used for a particular service which Digital Library.
That sounds good. You might also consider contacting the authors of the scales and getting their input on the work. They are probably interested in the same questions as you and may welcome the work to move the field forward. And they may be doing work not yet published that it will be useful to know about.
The attached article just out reviewing implementation science measures might be of use as they are focused on measures of similar constructs. On p. 6 and in references 18 & 19, there is discussion about measure validation criteria that might be useful.
- Badges
- Science method
- Similar topics
- Methodology
- Research Methodology
- Qualitative Research