Greetings Experts,
I am working on scale development. I started my data collection by using a 5-point Likert scale (strongly disagree to strongly agree) for my intended construct. After stage-I data collection, where i performed EFA on the dataset to assess the initial factor structure, I realised that to get more detailed results and to know more variations in the data I need high anchoring points so i shifted to 7 point Likert scale for my stage-II data collection, where i will confirm the initial factor solution derived through stage one data and run CFA to confirm the construct.
I need your guidance to know if I can use different anchoring points for Likert scale in EFA and CFA {perform EFA with 5 point scale, and CFA with 7 point scale}?
if yes, kindly provide any reference to support my decision, and if not, please guide how to proceed next, bcz I want to keep 7 point scale.
EFA and CFA are both based on the correlations among your items, so anything that affects those correlations (number of response categories, anchoring points, etc.) will affect the results of your factor analyses.
For EFA, be sure you avoid PCA and use something like Principal Factors or Maximum Likelihood instead, as well aa oblique rather than orthogonal rotation.
Just logically think, If you use both 5 and 7 point response for EFA and CFA respectively , what will be used in your recommended newly developed scale.
You didn't mention sample sizes of the two datasets. If they're sufficiently large, you can randomly split them, and run one subsample each of the 5-point group and the 7-point group with EFA and the other with CFA.
This (given adequate Ns for 5-point and 7-point) would either justify, refute or show neutrality to your 7-point preference. It would also suggest a strategy for other people who change their minds in midstream about how many response alternatives to use in a Likert scale. (Maybe, you could run an online gamble on which, if either, works best.)