If I have a set of traits that, based on standard epidemiological studies, are known to cluster in families (for example lymphomas and autoimmune diseases), but the genetic correlation matrix shows limited genetic correlation between lymphomas and autoimmune diseases, while showing high genetic correlations within each disease group, does it make sense to run a Genomic SEM including both autoimmune diseases and lymphomas? Or should one restrict the model to one disease group at a time?
Replying to “If I have a set of traits that, based on standard …”:
Based on the gen cor matrix it sounds like a (correlated) two-factor solution may make the most sense. You can also run EFA first to examine how these traits would cluster across two factors
Replying to “If I have a set of traits that, based on standard …”:
Yes I ran a EFA, that looked convincing for 2 factors. But when moving forward with two latent factors, the standard correlation between the two latent factors was -0.04. So now I am thinking that this makes little sense to move forward with. maybe I should just run the commonfactor() model for lymphomas (which are my main interest), because it does not seem to have much overlap with autoimmune diseases.
Replying to “If I have a set of traits that, based on standard …”:
It is already an interesting finding in itself that your two-factor solution that allows for correlated factors has a very low correlation between them! Im not sure how you were planning to move forward with the model but it is already an end point in itself :)