Thanks in advance!
Question: It is acceptable to run a measurement model using AMOS (and remove items) before testing my hypotheses using ANCOVA? Basically I want to examine the relationship between X (a dichotomous variable) and each dimension of multi-faceted job satisfaction which controlling for several variable (age, tenure, etc.). Given that the multi-faceted job satisfaction measure consist of 28 items, I want to run a measurement model to eliminate serious cross-loadings. The design is cross-sectional, not experimental.
My reviewer at a well-known journal indicates that running a CFA before ANCOVA is concerning because “this process treats the constructs as both formative and reflective constructs,” …. and “If the analysis is based on a formative construct that I do not need to conduct CFA”. The reviewer clearly indicates that they are not sure if their logic is right/wrong.
Any advice would be welcomed as would any cite.
My thought is that running a measurement model prior to performing regression/or structural model is encouraged. ANCOVA is a regression based technique. Given this, I believe I can run a CFA.
I have reviewed the literature extensively and still have not found an answer.