Hello, I need some advice regarding EFA and CFA, as follow:
1- Some studies say EFA and CFA can not been run the same data. In contrast, some others say there is no conflict between EFA and CFA to be conducted on the same data! Which one is more trustable.
2- Is EFA seriously needed for the standard scales that their translations are also available via prior studies?
I agree that EFA should be run for the modified scales, but not for the standard translated ones! What do you think?
3- In the cases that we are studying couples (husbands and wives separately), EFA and CFA should be conducted for them separately? If yes, what if the results of EFA for husbands and wives are found to be different?
Since the different categories in these spousal groups (regarding EFA) affect the sub-dimensions in CFA, how we should deal with it?
Please provide a reference for each part. Thank you so much for your valuable time.
Regards,
Hi,
if you have a clear assumption of the underlying structure you can use CFA directly. EFA is not neccessary then.
I dont know if I really get what you mean with question 3 and it would be more easy if one would know what model and data underlies your research. But if I get it right, I suggest to check for invariance between the groups (configural, metric, etc.)
Hope I could help!
Thank you for your response Mr. Vosylis.
Your explanation sounds logical. It was great if I can find an academic support for it.
Regards,
Thank you for your reply Mr. Utesch.
In my model, I studied the differences between husbands and wives in the level of each variables as one of my objectives. However, I run separate CFA, and SEM models for them.
Totally, I have six variables. I am done with data analysis. But, recently, I figured out that I have to run EFA as well. I run EFA for one one of the variables, the models were some how different in wives and husbands. I am wondering if the number of sub-dimensions and also, the items under each dimension are different, the shape of CFA models for husbands and wives should be drawn different as well. Additionally, the score should be computed again for husbands and wives because of different EFA models.
Hope I could say what I meant clearly.
Thank you,
EFA AND CFA are meant for two different purpose altogether.
1. EFA is meant to determine the dimensionality of items measuring the particular construct. To drop items which do not contribute to measuring the construct and also to assess the internal reliability of remaining items in measuring the construct. Here EFA is carried out based on pilot study data.
2. Once EFA is completed, the retained items will be used to measure the construct in term of measurement model. Using these items, the researcher collects the field study data (real study). Then, the CFA procedure is carried out to validate the measurement model for the construct.
3. The CFA should confirm the Validity and reliability of these constructs prior to modeling them in the Structural Model for estimating SEM.
4. Thus running EFA and CFA using the same data is totally wrong! Remember: Data should never be recycled.
Thank you for your reply Prof. Zainudin.
Would you please give your feedback regarding question 2 and 3 as well?
Regards,
For question no 2. EFA is not seriously needed but advisable to do since the original items have been translated into another language.
For question no 3.No problem if you obtain different results for two different genders.
Once again, thank you dear Prof.
So,
1- for instruments that have been used with no change and have already been translated in prior studies, there is no certain need for EFA. Am I right?
2- how do different EFA models for husbands and wives affect the other process of data analysis? I mean I should draw different CFA based on new factors obtained from EFA?
Thank you,
Best,
I suppose if for the male and female you have two different results, then you should take that into consideration when making your conclusions.
Thank you again Mr. Vosylis.
No! I am using SPSS for EFA and AMOS for CFA.
I have many variables. Until now, only one of them has been found to carry different factors in EFA among husbands (2 factors under one construct) and wives (3 factors under one construct). I guess in this case I should draw CFA for husband and wives differently! One with two sub-dimension, another one with three sub-dimension!
Thank you & Regards,
EFA is used to explore the factor structure among variables under study and CFA is used to confirm the factor structure which is already explored. initially data is collected for a set of variables and EFA is run. Based on the output results, variables with poor factor loadings and cross loadings are dropped and on left out variables CFA is run on the same data. Further, fresh data is collected for left out variables and CFA is run and model fit is checked based on goodness of fit indices, and convergent, divergent validates are checked.
- Badges
- Science topic