Model specification

[Xavier Farre]

Hello everyone,

I am still trying to figure out how to use genomicSEM, and learning by reading the wiki and all the issues reported here.

I have 6 traits and my idea was to obtain the GWAS of its common factor. But I saw something I can't understand while looking at the exploratory factor analysis.

Loadings:
  Factor1
T1          
T2      
T3 0.292  
T4 0.727  
T5 0.704  
T6 0.208  

               Factor1
SS loadings      1.163
Proportion Var   0.194


Loadings:
  Factor1 Factor2
T1 0.999
T2 0.383
T3 0.284        
T4 0.708        
T5 0.730        
T6 0.209        

               Factor1 Factor2
SS loadings      1.159   1.153
Proportion Var   0.193   0.192
Cumulative Var   0.193   0.385


  Factor1 Factor2 Factor3
T1 1.003                
T2 0.383                
T3 0.131 0.368
T4 0.656        
T5 0.819        
T6 0.748

               Factor1 Factor2 Factor3
SS loadings      1.161   1.126   0.700
Proportion Var   0.193   0.188   0.117
Cumulative Var   0.193   0.381   0.498

Why two of the traits don't appear if I select one factor?

Instead of using a model like this: F1 =~ T1 + T2 + T3 + T4 + T5 + T6

Would it be possible to use something like this?

F1 =~ T1 + T2

F2 =~ T3 + T4 + T5 + T6

L1 =~ F1 + F2

and test the SNP in L1?

I hope I explained myself properly.

Xavier

[agro...@gmail.com]

Hi Xavier, 

The psych package that you used to run the EFA doesn't print the loadings if they are below a certain threshold, which is why it does not print the loadings for T1 and T2 when you run a single-factor EFA. The results are also sensible given that T1 and T2 are the specific traits that load on the second factor when you expand to run a two-factor EFA. 

To your second question, it is certainly possible to test a hierarchical model like you've specified, and examine SNP effects on the hierarchical factor (L1). 

Best, 

  Andrew