After conducting EFA with the psych package and applying a CFA model in Lavaan, my fit indices and loadings seem inconsistent. Has anyone experienced these discrepancies?
hey id say its likely the default settings in lavaan vs psych causing the drop. slight differences in estimation and rotation details can mess with the numbers. check consistency in specs and factor retention, you might find the culprit.
hey i couldnt agree more. the intrinsic assumptions of each mdoel make a world of difference. lavaan in cfa enforces strict loadings while efa plays it loose; thus the diffrence in results. check your model specs for insights.
In my experience, the differences arise from the inherent approaches of these two methods. EFA is an exploratory analysis that accommodates cross-loadings and does not enforce strict relationships between variables. In contrast, CFA is a confirmatory analysis with a structured specification that limits flexibility and incorporates theory-driven constraints. Moreover, the implementation details vary between the psych and Lavaan packages. It is important to match the underlying assumptions and model specifications when comparing results from these two methods to avoid discrepancies.
Hey everyone, I’m really intrigued by this too! I’ve noticed similar discrepancies and it seems to boil down to the fact that EFA and CFA are really two different beasts. With EFA, you’re kind of letting the data play its own role, letting the factors emerge naturally which can be super flexible (and sometimes messy), whereas CFA ties everything down with predetermined paths based on your theory. One thing I came across is that even the rotation methods and how you handle estimation in each package can really change what you see. Have you looked into how different rotation or estimation options might be affecting your results? I’m curious to hear if anyone has experimented with tweaking these settings and what they found out. Any insights or experiences on this?