I am developing a questionnaire for social science. The preliminary questionnaire has 35 items with 6 domains. However, one of the domains was optional ( 5 items), where only those who had been involved in an accident before needed to answer that questions. Unfortunately, my answer option for answering the questionnaire did not include not applicable.
When I run the EFA, parallel analysis suggests 5 factors. But this optional domain loads closely with another item with different instructions/ other domains.
So my question is, can I remove this one domain ( 5 items) from the EFA analysis and run the remaining 30 items for EFA. Then, re-include this optional domain when I run for CFA. Can I use expert judgment /based on the importance of that optional domain to the questionnaire to retain it as it is?
( I try to rerun the 35 items and 6 domain/factor ( including the optional domain) using CFA, and the result indicated good convergent and discriminant validity)
I wonder if this method is permissible or how I should go about it? Thanks
Ms Khaireena When constructing a questionnaire, several dimensions or scales can be used for exploratory factor analysis (EFA) and confirmatory factor analysis (CFA). However, it is critical to evaluate the consequences of such a decision and to be open about the technique in your study report.
When utilizing EFA, it appears that the optional domain does not mesh well with the other questions in the questionnaire in your scenario. Removing it from the EFA analysis and then re-adding it in the CFA analysis might be a viable strategy, as long as the optional domain is still relevant to the overarching research topic and you have a strong rationale for including it in the CFA.
It's also worth noting that when you use CFA, you're testing a specific model, and this optional domain must load exactly as you described in your model. If it does not fit well in the model, it may impair the validity of your results, therefore keep that in mind while selecting whether to include or omit this domain.
Expert judgment can be used to support your decision to keep the optional domain, but it's also important to use other validation methods, such as cross-validating your results with other methods of analysis and performing sensitivity analysis to see how the results change with different sample sizes or model specifications.
To summarize, it is critical to be honest about your research methodologies and to examine the ramifications of deleting an optional domain from the EFA analysis while including it in the CFA analysis. Cross-validating your data with different techniques of analysis is also advised, as is consulting with an expert in SEM or a statistician to have a better understanding of the results.
EFA (Exploratory Factor Analysis) and CFA (Confirmatory Factor Analysis) are both statistical techniques used to analyze data from questionnaires, but they have some key differences.
In EFA, the goal is to identify underlying dimensions or factors in a dataset without preconceptions about the structure of the data. Therefore, the number and nature of the dimensions or factors are determined by the data. The researcher will have to decide on the number of factors to extract based on the scree plot and other factor extraction methods.
In CFA, the goal is to test a specific hypothesis about the structure of the data, usually based on a previous study or theory. Therefore, the researcher will have a preconceived idea about the number and nature of the dimensions or factors and will test it against the data.
In summary, EFA is used to identify underlying dimensions or factors in a dataset, whereas CFA is used to test a specific hypothesis about the structure of the data. Therefore, you may end up with different dimensions or scales when using EFA and CFA on the same questionnaire. It's important to keep in mind that the number of factors and their nature is different in EFA and CFA as it's based on different goals.
I do agree with the other two responses, and EFA is all about exploring the underlying dimensions in the indicators. So you can skip that optional domain given its nature and run it across the CFA. As you are aware of the importance of the expert validation, one more thing you can opt-in to validate the indicators used in that optional domain is by running it through the item analysis model. You can select classical test theory and see through the discrimination index and discriminatory power when you have collected the preliminary data (pilot-testing) and then can edit the indicators accordingly. And then, you can run the model in CFA alongside the data obtained after EFA to see through the model fit. But do mention all this in the methodology section and be transparent about it Ms Khaireena
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS