Factor loadings > 1 in CFA with lavaan

#1
Hi folks,

What are potential causes for factor loadings > 1 in CFA with lavaan? I'm using standardized loadings (std.lv). Is this generally concerning? Thanks
 

Lazar

Phineas Packard
#2
Factor loadings in a CFA with multiple correlated latent factors are akin to regression estimates not correlations, so having a loading over 1 is possible. Whether or not it is a concern is hard to say without more context.
 
#3
Factor loadings in a CFA with multiple correlated latent factors are akin to regression estimates not correlations, so having a loading over 1 is possible. Whether or not it is a concern is hard to say without more context.
Thanks. I'm cross validating two questionnaires and the > 1 loadings come from an established big five inventory. If loadings are akin to regression estimates, I suspect they might be able to amount to > 1 if the items are highly correlated?
 

Lazar

Phineas Packard
#4
Thanks. I'm cross validating two questionnaires and the > 1 loadings come from an established big five inventory. If loadings are akin to regression estimates, I suspect they might be able to amount to > 1 if the items are highly correlated?
Not quite. Remember the diagrams of a CFA. The arrows go from the latents to the items not the other way around. Hence, correlation between the items is not the issue here but the correlation between the latents.
 

Lazar

Phineas Packard
#5
In passing with measure of the big 5 you might want to try EFA or ESEM for more complicated models given known issues with model fit for the Big 5 in CFA models.
 
#6
Not quite. Remember the diagrams of a CFA. The arrows go from the latents to the items not the other way around. Hence, correlation between the items is not the issue here but the correlation between the latents.
Hm, but I also get factor loadings > 1 when specifying only one single factor.

In passing with measure of the big 5 you might want to try EFA or ESEM for more complicated models given known issues with model fit for the Big 5 in CFA models.
Never heard of ESEMs before, thanks!